A method for dividing coal spontaneous combustion three zones based on multilayer perceptron

Abstract

The present application belongs to the technical field of coal mine safety, and aims to solve the problem that the current commonly used oxygen volume fraction index division method is not accurate enough and is difficult to cope with complex coal mine environment and dynamic changes. A coal spontaneous combustion three-zone division method based on a multilayer perceptron is provided. The processed coal mine area feature data set is used as input data, a multilayer perceptron model is constructed, and the model is trained using the coal mine area feature data set. Then, real-time coal mine area feature data is input into the trained multilayer perceptron model to obtain the coal spontaneous combustion three-zone area division result. In a complex data environment, the coal spontaneous combustion three-zone area can be efficiently identified, the spontaneous combustion risk assessment can be performed more quickly, and an intuitive division result can be provided, which is convenient for on-site decision-making and reduces errors caused by human participation.

Description

Technical Field

[0001] This invention relates to the field of coal mine safety technology, specifically to a method for classifying the three zones of spontaneous combustion in coal based on a multilayer sensor. Background Technology

[0002] Coal resources still occupy a major position in my country's resource structure, but spontaneous combustion of coal is a common safety hazard in coal mining. Spontaneous combustion occurs because, during the mining process, oxygen enters the goaf through mining-induced cracks, causing residual coal to accumulate heat, oxidize, and spontaneously combust. To prevent spontaneous combustion of residual coal in the goaf, it is necessary to delineate a heat dissipation zone, an oxidation zone, and an asphyxiation zone—the so-called "three zones" of spontaneous combustion.

[0003] Traditional methods for delineating spontaneous combustion zones rely heavily on the experience of underground workers, using their sense of smell to identify potential coal seam spontaneous combustion areas, or employing simple physical models to coarsely delineate goaf areas, resulting in overly broad delineation. Currently, the oxygen volume fraction index is commonly used, but this method is not accurate enough and struggles to cope with the complex and dynamic nature of coal mine environments. Therefore, there is an urgent need to develop a method capable of efficiently delineating the three spontaneous combustion zones in complex environments, thereby effectively preventing coal seam spontaneous combustion. Summary of the Invention

[0004] To solve the above-mentioned technical problems, or at least partially solve them, this disclosure provides a method for classifying the three zones of spontaneous combustion in coal based on a multilayer sensor.

[0005] This invention provides a method for classifying three zones of spontaneous combustion in coal based on a multilayer sensor. The method includes: acquiring a coal mine area feature dataset, and preprocessing the coal mine area feature dataset to obtain standard feature data. The coal mine area feature data includes at least temperature, oxygen concentration, carbon monoxide concentration, and air leakage.

[0006] A multilayer perceptron model is constructed, with standard feature data as the input to the model. The probability division results of the three zones of spontaneous combustion in coal are obtained from the multilayer perceptron model. The multilayer perceptron model is trained and validated based on the above coal mine regional feature data.

[0007] Real-time coal mine regional feature data are input into a trained multilayer perceptron model to obtain the three-zone division results of coal spontaneous combustion.

[0008] Preferably, the multilayer perceptron model includes at least an input layer, two hidden layers, and an output layer. The input layer has 4 input nodes, corresponding to temperature, oxygen concentration, carbon monoxide concentration, and air leakage. The first and second hidden layers are used to extract high-level representations from the coal mine area feature data. The output layer has 3 neurons, corresponding to the heat dissipation zone, oxidation zone, and asphyxiation zone in the three zones.

[0009] The optimal multilayer perceptron model outputs the probability distributions of the heat dissipation zone, oxidation zone, and suffocation zone, and selects the zone with the highest probability as the final output zone of the model.

[0010] Preferably, if P(heat dissipation zone) > P(oxidation zone) and P(heat dissipation zone) > P(suffocation zone), then the zone is determined to be a heat dissipation zone;

[0011] If P(oxidation zone) > P(heat dissipation zone) and P(oxidation zone) > P(asphyxiation zone), then the zone is determined to be an oxidation zone.

[0012] If P(asphyxiation zone) > P(heat dissipation zone) and P(asphyxiation zone) > P(oxidation zone), then the zone is determined to be an asphyxiation zone.

[0013] The preferred method involves preprocessing the coal mine area feature dataset, specifically through the following steps:

[0014] The mean imputation method was used to process the regional feature data of the coal mine, replacing the missing values.

[0015] Use box plots to identify and process outliers in coal mine area feature data;

[0016] The Gaussian kernel density function is used to estimate the coal mine area feature data after the above processing, and then the estimation results are subjected to logarithmic transformation;

[0017] The logarithmic transformation result is subjected to min-max normalization to obtain standard feature data.

[0018] The preferred method involves training the multilayer perceptron model using the training set, specifically including:

[0019] The difference between the predicted probability distribution and the true probability distribution is calculated using a loss function to evaluate the prediction accuracy of the multilayer perceptron model; an optimizer is used to process the coal mine area feature data and automatically adjust the learning speed of the multilayer perceptron model; the training set is divided into several smaller training sets, and the parameters of the multilayer perceptron model are updated batch by batch using mini-batch gradient descent.

[0020] The preferred approach is to divide the monitored coal mine area feature dataset into a training set and a validation set. Specifically, the K-fold cross-validation method is used to divide the coal mine area feature dataset into K subsets, and each subset is used as the validation set, while the rest are used as the training set.

[0021] After selecting and obtaining the trained multilayer perceptron model, the process also includes: evaluating the multilayer perceptron model according to preset indicators, and adjusting and optimizing the multilayer perceptron model based on the evaluation results.

[0022] The optimized preset metrics include, but are not limited to, accuracy, recall, precision, and F1 score; accuracy is the proportion of correctly predicted samples, recall is the proportion of correctly identified positive samples, precision is the proportion of predicted positive samples that are actually positive samples, and the F1 score is the harmonic mean of precision and recall.

[0023]

[0024] In the formula, TP and TN represent the number of correctly predicted positive classes and the number of correctly predicted negative classes, respectively; FP and FN represent the number of incorrectly predicted positive classes and the number of incorrectly predicted negative classes, respectively.

[0025] The preferred positive class represents the target that the multilayer perception model focuses on. In the classification task, each time a specific region is predicted, that region is the positive class, and other regions are the negative class. The positive and negative classes change dynamically according to the current prediction target.

[0026] The technical solution provided in this disclosure has the following advantages compared with the prior art:

[0027] This invention constructs a multilayer perceptron model by using the processed coal mine area feature dataset as input data, and trains the model with data. The trained multilayer perceptron model can efficiently identify the three zones of spontaneous combustion of coal in complex data environments, thereby enabling faster spontaneous combustion risk assessment and providing intuitive classification results to facilitate on-site decision-making. Moreover, data training can also improve the accuracy and robustness of the model and reduce errors caused by human intervention. Attached Figure Description

[0028] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0029] Figure 1 This is a flowchart illustrating the coal spontaneous combustion three-zone division method based on a multilayer sensor according to an embodiment of the present invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] The division of coal spontaneous combustion zones into three areas is based on the existence of spontaneous combustion zones. According to the theory of ventilation and oxygen supply and coal combustion composites, the goaf is divided into different zones: the heat dissipation zone, the oxidation zone, and the asphyxiation zone. The heat dissipation zone is close to the working face, where residual coal will not spontaneously combust. The oxidation zone is adjacent to the heat dissipation zone, suitable for coal oxidation reactions that ultimately lead to spontaneous combustion. The asphyxiation zone is located at the deepest part of the goaf, with a very low oxygen concentration, failing to meet the conditions required for residual coal to spontaneously combust, and therefore, spontaneous combustion will not occur there. Therefore, it is necessary to accurately determine the division of the three coal spontaneous combustion zones to facilitate the implementation of corresponding measures to prevent and control spontaneous combustion and ensure coal mine safety.

[0032] Currently, the classification of the three zones of spontaneous combustion in coal mines mainly relies on the experience of underground workers, using their sense of smell to identify potential spontaneous combustion areas, or using simple physical models to coarsely divide the goaf into overly broad areas. The commonly used method is based on oxygen volume fraction, but this method is not accurate enough and struggles to handle the complex and dynamic nature of coal mine environments. This invention provides a method for classifying the three zones of spontaneous combustion in coal mines based on a multilayer sensor. This method can accurately identify the three zones of spontaneous combustion in complex data environments, providing intuitive classification results and enabling more accurate assessment of spontaneous combustion risks.

[0033] This embodiment provides a method for classifying the three zones of spontaneous combustion in coal based on a multilayer sensor, which can be used in the aforementioned mobile terminals, such as all-in-one computers and tablet computers. Figure 1 This is a flowchart of a coal spontaneous combustion three-zone division method based on a multilayer sensor according to an embodiment of the present invention, as shown below. Figure 1 As shown, the process includes the following steps:

[0034] Step S100: Obtain a coal mine area feature dataset and preprocess the coal mine area feature dataset to obtain standard feature data. The coal mine area feature data includes at least temperature, oxygen concentration, carbon monoxide concentration, and air leakage.

[0035] In this embodiment, monitoring points are set up in the coal mine goaf area. Each monitoring point is equipped with a temperature sensor, an oxygen sensor, a carbon monoxide sensor, and an anemometer to obtain four key characteristic data of the coal mine goaf area: temperature, oxygen concentration, carbon monoxide concentration, and air leakage.

[0036] Optionally, the coal mine area feature dataset is preprocessed, specifically by: using mean imputation to replace missing values ​​in the coal mine area feature data; using box plots to identify and process outliers in the coal mine area feature data; using Gaussian kernel density function to estimate the coal mine area feature data after the above processing, and then performing a logarithmic transformation on the estimation result; and performing max-min normalization on the logarithmic transformation result to obtain standard feature data.

[0037] In this embodiment, the mean imputation method uses multi-feature mean imputation to obtain the features of each data in the coal mine area feature dataset, calculate the correlation between each feature and other features, determine the weighting coefficient based on the correlation, and then use the weighted mean to imput the missing values, so that the imputed data is more consistent with the actual working conditions of the coal mine.

[0038] This invention utilizes the relationships between multiple features in a coal mine regional feature dataset to calculate a weighted mean, enabling more accurate filling of missing values. For example, in practical applications, missing values ​​can be filled based on the correlation between oxygen concentration and temperature, combined with the weighted mean of the features of their respective data, making the processed data more consistent with actual coal mine working conditions.

[0039] In this embodiment, the mean imputation method is used to process missing values ​​in the coal mine area feature dataset, replacing the missing values ​​in the dataset with the mean of the corresponding data. Then, box plots are used to identify and process outliers in the dataset, and the quartiles of the data are calculated, including the lower quartile (Q1) and the upper quartile (Q3). The interquartile range (IQR) is then calculated to determine the outlier limits. The lower limit is calculated as Q1 - 1.5 × IQR and the upper limit is calculated as Q3 + 1.5 × IQR. Values ​​in the data that exceed these limits are identified as outliers, and these outliers are deleted or adjusted as needed.

[0040] In this embodiment, the Gaussian kernel density function formula is: Where n is the sample size; h is the bandwidth; x is the evaluation point; x i denoted as data points; K is the Gaussian kernel function.

[0041] In this embodiment, the logarithmic transformation formula is: y = log(x + c), where c is an adjustment coefficient. In practical applications, c is a preset coefficient, which can be 1 × 10⁻⁶. -10 As long as it can be ensured that the data will not have the problem of undefined logarithms during transformation, this invention is not limited thereto.

[0042] In this embodiment, the data is processed using max-min normalization, scaling the data to the [0,1] interval to unify the feature scale. The max-min normalization formula is: Where x is the normalized data; x' is the normalized data; xmin It is the minimum value in the dataset; x max It is the maximum value in the dataset.

[0043] Step S200: Construct a multilayer perceptron model, which includes at least an input layer, two hidden layers, and an output layer. The input layer is used to receive the standard feature data, the hidden layers are used to process the standard feature data, and the output layer is used to determine the probability division result of the three zones of spontaneous combustion of coal based on the data processing result of the hidden layers.

[0044] Optionally, a multilayer perceptron model is constructed, specifically including: the input nodes of the input layer are set to 4, corresponding to temperature, oxygen concentration, carbon monoxide concentration and air leakage; the number of neurons in the first hidden layer is set to 32, and the number of neurons in the second hidden layer is set to 16, which are used to extract high-level representations from the feature data of the coal mine area; the number of neurons in the output layer is set to 3, corresponding to the heat dissipation zone, oxidation zone and asphyxiation zone in the three zones.

[0045] In this embodiment, the multilayer perceptron model consists of an input layer, two hidden layers, and an output layer. The input layer receives standard feature data with 4 input nodes, corresponding to temperature, oxygen concentration, carbon monoxide concentration, and air leakage, respectively. The first hidden layer has 32 neurons, and the second hidden layer has 16 neurons, further extracting high-level representations of the features. The ReLU activation function is used to capture the nonlinear relationships of the features. The output layer has 3 neurons, corresponding to the heat dissipation zone, oxidation zone, and asphyxiation zone, respectively. The Softmax activation function is used to output the probability of each category, transforming the output into a probability distribution to achieve probability division of the three zones.

[0046] In this embodiment, the high-level representation formula is x = [x1, x2, x3, x4], where x1 is temperature, x2 is oxygen concentration, x3 is carbon monoxide concentration, and x4 is air leakage. The first hidden layer has 32 neurons. Each neuron uses a weight matrix W1 and a bias vector b1 to perform a weighted summation of the input, and then applies the ReLU activation function to capture the nonlinear relationships between the features. The weighted summation formula is: h1 = ReLU(W1·x + b1), where h1 is the output of the first hidden layer, i.e., the high-level representation of the first hidden layer; W1 is a weight matrix of size 32×4; and b1 is a bias vector of size 32×1.

[0047] In this embodiment, the formula for the ReLU activation function is: f(x) = max(0,x), where f(x) is the output after activation and x represents the input value.

[0048] In this embodiment, the second hidden layer has 16 neurons and performs further feature extraction on the output of the first hidden layer. The formula is: h2=RELU(W2·h1+b2), where h2 is the output of the second hidden layer, that is, the high-level representation of the second hidden layer; W2 is a weight matrix of size 16×4; and b2 is a bias vector of size 16×1.

[0049] In this embodiment, the output layer has 3 neurons, corresponding to the heat dissipation zone, oxidation zone, and asphyxiation zone, and its calculation formula is as follows:

[0050] z = W3·h2 + b3, where z is the original score of the output layer, i.e., logits; W3 is a 3×16 weight matrix; and b3 is a 3×1 bias vector. After the logits of the output layer are processed by the Softmax activation function, the probabilities corresponding to the heat dissipation zone, oxidation zone, and suffocation zone are output.

[0051] In this embodiment, each neuron outputs a logits, representing the activation intensity of that class. To convert these scores into probabilities, a Softmax activation function is used. The output probability distribution is then used, and the band with the highest probability is selected as the final output band. The model's output logits are converted into three probability values: P(heat dissipation zone), P(oxidation zone), and P(asphyxiation zone), the sum of which is 1. The model outputs a probability value belonging to each spontaneous combustion zone, explicitly indicating which spontaneous combustion zone each data point or coal mine area belongs to. The specific determination principle is as follows:

[0052] If P(heat dissipation zone) > P(oxidation zone) and P(heat dissipation zone) > P(suffocation zone), then the location is determined to be a heat dissipation zone.

[0053] If P(oxidation zone) > P(heat dissipation zone) and P(oxidation zone) > P(suffocation zone), then the location is determined to be an oxidation zone.

[0054] If P(asphyxiation zone) > P(heat dissipation zone) and P(asphyxiation zone) > P(oxidation zone), then the location is determined to be an asphyxiation zone.

[0055] In this embodiment, the formula for the Softmax activation function is: Wherein, P(y i ) represents the probability of the i-th class, i.e., the probability of a certain spontaneous combustion zone; z i It is the logits corresponding to the i-th class; This indicates that there are three categories, corresponding to the heat dissipation zone, oxidation zone, and suffocation zone, respectively.

[0056] In this embodiment, the nonlinear relationships of features captured by the ReLU activation function include nonlinear feature mappings, such as: 1. Nonlinear transformation of linear combinations: Neurons in each layer process the input through weighted sums and activation functions. 2. Feature interactions: Through a multi-layer neural network structure, the multilayer perceptron model can automatically capture the complex interactive relationships between features.

[0057] In this embodiment, the nonlinear relationship between temperature, oxygen concentration, carbon monoxide concentration, and air leakage can be learned through nonlinear mapping of hidden layer neurons. Specifically:

[0058] Nonlinear Relationship between Temperature and Oxygen Concentration: In the spontaneous combustion zone of coal mines, a significant nonlinear relationship exists between temperature and oxygen concentration, with the accelerating effect of temperature increase. Typically, as temperature rises, the rate of oxygen consumption accelerates, leading to the formation of oxidation and asphyxiation zones. Using the ReLU activation function, a multilayer perceptron model can learn a nonlinear pattern where oxygen consumption is slower at lower temperatures, but the oxygen concentration drops sharply after the temperature rises to a certain threshold.

[0059] The nonlinear relationship between temperature and carbon monoxide concentration: Carbon monoxide is an early indicator gas in the spontaneous combustion process. In the initial stage of coal spontaneous combustion (i.e., the oxidation zone), carbon monoxide concentration gradually increases. With further increases in temperature, the amount of carbon monoxide generated increases rapidly. Multilayer perceptron models, through nonlinear activation functions, can capture the complex relationship between carbon monoxide concentration and temperature. (For example, at lower coal temperatures, carbon monoxide concentration is not significant, but when the temperature approaches the spontaneous combustion critical temperature, the carbon monoxide concentration increases sharply, forming a nonlinear "peak effect.")

[0060] Multi-feature interactive nonlinear relationships: the combined effect of temperature, oxygen concentration, and carbon monoxide concentration. Increased temperature accelerates the oxidation reaction, leading to a decrease in oxygen concentration and an increase in carbon monoxide concentration; conversely, a decrease in oxygen concentration inhibits further intensification of the oxidation reaction. Multilayer perceptron (MLP) models, through the connection weights of neurons in the hidden layers and the ReLU activation function, can capture this complex interaction, helping MLP models determine the progress of spontaneous combustion.

[0061] The coupling effect between temperature and air leakage: Increased coal seam temperature accelerates the spontaneous combustion reaction, while changes in air leakage affect heat dissipation and oxygen supply, leading to varying spontaneous combustion rates in different zones. Multilayer perceptron models can identify the impact of air leakage on the formation of different spontaneous combustion zones in coal seams through this nonlinear relationship.

[0062] Nonlinear distribution characteristics of the three zones of spontaneous combustion in coal: Nonlinear boundary between heat dissipation zone, oxidation zone and asphyxiation zone: As the coal temperature increases, the spontaneous combustion zone transitions from the heat dissipation zone to the oxidation zone and then to the asphyxiation zone. The multilayer perceptron model can capture the complex transformation process between these three zones through nonlinear transformation of multiple hidden layers, and output the probability of each zone through the final Softmax layer.

[0063] Step S300: Divide the coal mine area feature dataset into a training set and a validation set. Use the training set to train the multilayer perceptron model, and then use the validation set to determine the optimal parameters to obtain the trained multilayer perceptron model.

[0064] In this embodiment, the relevant parameters for training the multilayer perceptron model can be set as follows: learning rate of 0.001, batch size of 32, and number of training epochs of 50. L2 regularization is used to prevent overfitting, and the Xavier initialization method is used to ensure reasonable initial weights.

[0065] In this embodiment, the coal mine area feature dataset is divided into a training set and a validation set according to 80% and 20% respectively. Random noise is added to the data training set, and a Gaussian noise filter is used to smooth the data.

[0066] Optionally, the training set can be divided into several smaller training sets, and the parameters of the multilayer perceptron model can be updated batch by batch using mini-batch gradient descent.

[0067] Optionally, the coal mine area feature dataset can be divided into K subsets using the K-fold cross-validation method, with each subset used as the validation set and the remainder used as the training set.

[0068] Optionally, the optimal parameters can be determined using the validation set, specifically by using the accuracy, recall, precision, and F1 score of the validation set.

[0069] Optionally, the multilayer perceptron model is trained using the training set, specifically including: using a loss function to calculate the difference between the predicted probability distribution and the true probability distribution to evaluate the prediction accuracy of the multilayer perceptron model; and using an optimizer to process the coal mine area feature data and automatically adjust the learning speed of the multilayer perceptron model.

[0070] In this embodiment, the optimizer can be the Adam optimizer, which converges quickly due to its adaptive learning rate characteristic, making it suitable for processing such complex data. The optimizer update rule is: m t =β1m t-1 +(1-β1)g t ,

[0071] Where, m t V is the first-order momentum of the gradient;t Let g be the second momentum of the gradient; β1 and β2 are the exponential decay rates of the first and second momentum, respectively, and are taken as 0.9 and 0.999; g t The current gradient; and These are the momentum and second moment after bias correction, respectively; α is the learning rate, taken as 0.001; θ t For the updated parameters, θ represents all trainable parameters in the multilayer perceptron model, that is, the set of weight matrices and bias vectors of all layers in the multilayer perceptron model.

[0072] Step S400: Input the real-time coal mine area feature data into the trained multilayer perceptron model to obtain the coal spontaneous combustion three-zone division result.

[0073] This invention uses a processed coal mine area feature dataset as input data to construct a multilayer perceptron model, and trains the model with the data. The trained multilayer perceptron model can efficiently identify the three zones of spontaneous combustion of coal in complex data environments, thereby enabling faster spontaneous combustion risk assessment and providing intuitive classification results for easier on-site decision-making. Moreover, data training also improves the accuracy and robustness of the model, reducing errors caused by human intervention.

[0074] Step S500: Evaluate the multi-layer perception model according to preset indicators, and adjust and optimize the multi-layer perception model based on the evaluation results.

[0075] In this embodiment, the performance of the multilayer perception model is evaluated using preset metrics such as accuracy, recall, precision, and F1 score, and the model is then adjusted and optimized.

[0076] In this embodiment, accuracy is the proportion of correctly predicted samples, recall is the proportion of correctly identified positive samples, precision is the proportion of samples predicted as positive that were actually positive, and the F1 score is the harmonic mean of precision and recall.

[0077]

[0078] In the formula, TP and TN represent the number of correctly predicted positive classes and the number of correctly predicted negative classes, respectively; FP and FN represent the number of incorrectly predicted positive classes and the number of incorrectly predicted negative classes, respectively.

[0079] In this embodiment, the positive class represents the target that the multilayer perceptron model focuses on, that is, the specific category it wants to identify. In the classification task, each time a specific region is predicted, that region is the positive class, and other regions are the negative class. The positive and negative classes change dynamically according to the current prediction target.

[0080] For example, if the focus is on predicting the oxidation zone, then the oxidation zone is considered positive, and the other two zones are considered negative. If the focus is on the heat dissipation zone, then the heat dissipation zone is considered positive, and the other two zones are considered negative. Suppose we are currently focusing on the oxidation zone: then the oxidation zone is considered positive, and the other two zones are considered negative.

[0081] TP: The multilayer sensing model predicts that the sample is a positive class (oxidation zone), and the sample is indeed a positive class (located in the oxidation zone).

[0082] TN: The multilayer sensing model predicts the sample as a negative class (i.e., the heat dissipation zone or the suffocation zone), and the sample is actually also a negative class (located in the heat dissipation zone or the suffocation zone).

[0083] FP: The multilayer sensing model predicts the sample as a positive class (oxidation zone), but the sample is actually a negative class (i.e., it is actually a heat dissipation zone or a suffocation zone).

[0084] FN: The multilayer sensing model predicts the sample as a negative class (i.e., the heat dissipation zone or the suffocation zone), but the sample is actually a positive class (i.e., the actual oxidation zone).

[0085] In this embodiment of the invention, the trained multilayer perceptron model is evaluated by preset indicators, and the constructed model is further optimized and adjusted so that the multilayer perceptron model can more accurately obtain the division results of the three zones of spontaneous combustion of coal under complex environments.

[0086] To verify the effectiveness and feasibility of the method of the present invention, an experiment was conducted using data collected from a coal mine area as an example.

[0087] The collected data included: temperature (T): 50℃; oxygen concentration (O2): 16%; carbon monoxide concentration (CO): 0.05%; air leakage (Q): 12m³. 3 / min. Using the multilayer perceptron model designed in this invention, the first hidden layer extracts the following combinations of nonlinear features:

[0088] Feature 1: As temperature increases, oxygen concentration decreases rapidly (non-linear relationship).

[0089] Feature 2: The carbon monoxide concentration increases non-linearly with increasing temperature and air leakage.

[0090] Feature 3: The critical transition point between the oxidation zone and the asphyxiation zone at a certain temperature.

[0091] The above features are further combined in the second hidden layer, and the final output probability distribution is: heat dissipation zone: 0.1; oxidation zone: 0.7; suffocation zone: 0.2.

[0092] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for classifying three zones of spontaneous combustion in coal based on a multilayer sensor, characterized in that, The method includes: A coal mine area feature dataset is obtained, and the coal mine area feature dataset is preprocessed to obtain standard feature data. The coal mine area feature data includes at least temperature, oxygen concentration, carbon monoxide concentration, and air leakage. The coal mine area feature dataset is preprocessed, specifically through the following steps: The mean imputation method was used to process the regional feature data of the coal mine, replacing the missing values. Use box plots to identify and process outliers in coal mine area feature data; The Gaussian kernel density function is used to estimate the coal mine area feature data after the above processing, and then the estimation results are subjected to logarithmic transformation; The logarithmic transformation result is subjected to min-max normalization to obtain standard feature data; A multilayer perceptron model is constructed, using standard feature data as input. The multilayer perceptron model includes at least an input layer, two hidden layers, and an output layer. The input layer receives the standard feature data, the hidden layers process the standard feature data, and the output layer determines the probability division of the three zones of spontaneous combustion of coal based on the data processing results of the hidden layers. The multilayer perceptron model is trained and validated based on the training set composed of the aforementioned coal mine area feature data. Real-time coal mine regional feature data are input into a trained multilayer perceptron model to obtain the three-zone division results of coal spontaneous combustion.

2. The method for dividing coal spontaneous combustion zones based on a multilayer sensor according to claim 1, characterized in that: The multilayer perceptron model includes at least an input layer, two hidden layers, and an output layer. The input layer has 4 input nodes, corresponding to temperature, oxygen concentration, carbon monoxide concentration, and air leakage. The first and second hidden layers are used to extract high-level representations from the feature data of the coal mine area. The output layer has 3 neurons, corresponding to the heat dissipation zone, oxidation zone, and asphyxiation zone in the three zones.

3. The method for dividing coal spontaneous combustion zones based on a multilayer sensor according to claim 2, characterized in that: The multilayer perceptron model outputs the probability distributions of the heat dissipation zone, oxidation zone, and suffocation zone, and selects the zone with the highest probability as the final output zone of the model.

4. The method for dividing coal spontaneous combustion zones based on a multilayer sensor according to claim 3, characterized in that: If P(heat dissipation zone) > P(oxidation zone) and P(heat dissipation zone) > P(suffocation zone), then the final output zone is determined to belong to the heat dissipation zone. If P(oxidation band) > P(heat dissipation band) and P(oxidation band) > P(suffocation band), then the final output band is determined to belong to the oxidation band. If P(suffocation zone) > P(heat dissipation zone) and P(suffocation zone) > P(oxidation zone), then the final output zone is determined to be a suffocation zone.

5. The method for dividing coal spontaneous combustion zones based on a multilayer sensor according to claim 1, characterized in that: Training the multilayer perceptron model using the training set specifically includes: The difference between the predicted probability distribution and the true probability distribution is calculated using a loss function to evaluate the prediction accuracy of the multilayer perceptron model; an optimizer is used to process the coal mine area feature data and automatically adjust the learning speed of the multilayer perceptron model; the training set is divided into several smaller training sets, and the parameters of the multilayer perceptron model are updated batch by batch using mini-batch gradient descent.

6. The method for dividing coal spontaneous combustion zones based on a multilayer sensor according to claim 1, characterized in that: The monitored coal mine area feature dataset is divided into a training set and a validation set, specifically including: The coal mine area feature dataset is divided into K subsets using K-fold cross-validation. Each subset is used as the validation set, and the remainder is used as the training set. The accuracy, recall, precision, and total precision of the validation set are then used as the training set. F One score determines the optimal parameters of the multilayer perceptron model.

7. The method for dividing coal spontaneous combustion zones based on a multilayer sensor according to claim 1, characterized in that: After obtaining the trained multilayer perceptron model, the process also includes: evaluating the multilayer perceptron model according to preset metrics, and adjusting and optimizing the multilayer perceptron model based on the evaluation results.

8. The method for dividing coal spontaneous combustion zones based on a multilayer sensor according to claim 7, characterized in that: Preset metrics include accuracy, recall, precision, and Accuracy is the proportion of correctly predicted samples; recall is the proportion of correctly identified positive samples; precision is the proportion of samples predicted as positive that were actually positive. The harmonic mean of precision and recall is given by: ， ， ， ， In the formula, TP and TN These represent the number of correctly predicted positive classes and the number of correctly predicted negative classes, respectively. FP and FN These represent the number of incorrectly predicted positive classes and the number of incorrectly predicted negative classes, respectively.

9. The method for dividing coal spontaneous combustion zones based on a multilayer sensor according to claim 8, characterized in that: Positive classes represent the targets that the multilayer perceptron model focuses on. In classification tasks, each time a specific region is predicted, that region is the positive class, and other regions are the negative classes. The positive and negative classes change dynamically according to the current prediction target.

Images