CN116975775B - Deep learning method for coal mine gas outburst prediction based on multi-mode data fusion - Google Patents
Deep learning method for coal mine gas outburst prediction based on multi-mode data fusion Download PDFInfo
- Publication number
- CN116975775B CN116975775B CN202310790338.0A CN202310790338A CN116975775B CN 116975775 B CN116975775 B CN 116975775B CN 202310790338 A CN202310790338 A CN 202310790338A CN 116975775 B CN116975775 B CN 116975775B
- Authority
- CN
- China
- Prior art keywords
- data
- gas outburst
- gas
- prediction
- coal mine
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 239000003245 coal Substances 0.000 title claims abstract description 56
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000013135 deep learning Methods 0.000 title claims abstract description 20
- 230000004927 fusion Effects 0.000 title claims abstract description 17
- 238000007781 pre-processing Methods 0.000 claims abstract description 14
- 230000006870 function Effects 0.000 claims description 30
- 230000008569 process Effects 0.000 claims description 16
- 238000000605 extraction Methods 0.000 claims description 13
- 238000010606 normalization Methods 0.000 claims description 12
- 238000011176 pooling Methods 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 5
- 235000004257 Cordia myxa Nutrition 0.000 claims description 4
- 244000157795 Cordia myxa Species 0.000 claims description 4
- 230000002159 abnormal effect Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 4
- 230000003213 activating effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000009194 climbing Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000009423 ventilation Methods 0.000 description 1
Classifications
-
- E—FIXED CONSTRUCTIONS
- E21—EARTH OR ROCK DRILLING; MINING
- E21F—SAFETY DEVICES, TRANSPORT, FILLING-UP, RESCUE, VENTILATION, OR DRAINING IN OR OF MINES OR TUNNELS
- E21F17/00—Methods or devices for use in mines or tunnels, not covered elsewhere
- E21F17/18—Special adaptations of signalling or alarm devices
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
- G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/27—Regression, e.g. linear or logistic regression
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/0985—Hyperparameter optimisation; Meta-learning; Learning-to-learn
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/02—Agriculture; Fishing; Forestry; Mining
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Mining & Mineral Resources (AREA)
- Business, Economics & Management (AREA)
- Probability & Statistics with Applications (AREA)
- Animal Husbandry (AREA)
- Human Resources & Organizations (AREA)
- Agronomy & Crop Science (AREA)
- Geochemistry & Mineralogy (AREA)
- Marine Sciences & Fisheries (AREA)
- Geology (AREA)
- Economics (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a coal mine gas outburst prediction deep learning method based on multi-mode data fusion, which comprises the following steps of S1: different sensors acquire a plurality of gas outburst data under different gas states; the gas outburst data comprise acoustic emission data, gas pressure data and coal resistivity data; step S2: carrying out data preprocessing on the gas outburst data; step S3: a depth prediction network for coal mine gas outburst prediction is constructed, and gas outburst data after data preprocessing is input; step S4: training iteration of a depth prediction network for coal mine gas outburst prediction is executed to form a depth prediction model; step S5: and predicting the probability of occurrence of coal mine gas outburst by using the depth prediction model to obtain a gas outburst prediction result.
Description
Technical Field
The invention relates to a gas outburst prediction method, in particular to a coal mine gas outburst prediction deep learning method based on multi-mode data fusion.
Background
When the underground gas outburst of the coal mine occurs, a large amount of gas and crushed coal are sprayed out of the coal seam in a very short time, so that holes are formed in the coal seam, a ventilation system or tunnel facilities of the coal mine are easily damaged, a large amount of miners are buried due to the fact that the coal mine collapses more seriously, and the safety of coal mining is seriously threatened. Therefore, if the intensity of the gas outburst can be rapidly and accurately predicted before the gas outburst accident occurs, the accident can be prevented from occurring, and damage and loss to coal production can be prevented. For gas outburst prediction, researchers have proposed various methods such as: patent CN112183901a proposes a "method for predicting the intensity of coal and gas protrusion based on deep learning", and patent CN109711632a proposes a "method for predicting the intensity of coal and gas protrusion based on abnormal sensitivity index of gas emission". The method based on deep learning is adopted to improve the prediction accuracy, but the gas outburst prediction is limited due to single acquired information; the gas outburst is predicted by three indexes of abnormal average fluctuation amplitude change of gas emission, abnormal average change trend of gas emission and abnormal frequency of large amplitude change of gas emission, and the risk of gas outburst can be predicted in real time, but the accuracy is slightly poor.
Disclosure of Invention
In order to solve the problems, the invention aims to provide a coal mine gas outburst prediction deep learning method based on multi-mode data fusion, namely multi-branch and attention mechanisms are adopted in a deep convolution network to fully utilize multi-source data information, so that the robustness of the method and the accuracy of gas outburst prediction are improved.
The invention discloses a coal mine gas outburst prediction deep learning method based on multi-mode data fusion, which comprises the following steps of S1: different sensors acquire a plurality of gas outburst data under different gas states; the gas outburst data comprise acoustic emission data, gas pressure data and coal resistivity data;
step S2: carrying out data preprocessing on the gas outburst data;
step S3: a depth prediction network for coal mine gas outburst prediction is constructed, and gas outburst data after data preprocessing is input;
step S4: training iteration of a depth prediction network for coal mine gas outburst prediction is executed to form a depth prediction model;
step S5: and predicting the probability of occurrence of coal mine gas outburst by using the depth prediction model to obtain a gas outburst prediction result.
Further, step S2 includes the steps of:
step S21: respectively carrying out missing value processing on acoustic emission data, gas pressure data and coal resistivity data of the gas outburst data;
step S22: and respectively carrying out normalization processing on the acoustic emission data, the gas pressure data and the coal resistivity data of the gas outburst data subjected to the missing value processing.
In step S21, the missing value processing is to delete the data with more missing values of the gas projection, and the data with less missing values of the gas projection is interpolated by KNN algorithm;
in step S22, the normalization process is to restrict the missing value interpolated gas projection data to the range of [0,1] by normalization.
Further, the depth prediction network comprises a feature extraction module, a block attention module and a regression network module which are sequentially connected; the characteristic extraction module comprises three branches, and the three branches respectively correspond to acoustic emission data, gas pressure data and coal resistivity data for processing the gas outburst data; each branch comprises a depth gating convolutional network module and a channel attention module; the block attention module fuses the features output by the three branches of the feature extraction module.
Further, in step S3, the data preprocessing process of the gas highlighting data is as follows:
step S31: respectively inputting acoustic emission data, gas pressure data and coal resistivity data of the gas outburst data after data preprocessing into a depth gating convolution network module and a channel attention module of corresponding branches to perform feature extraction;
step S32: fusing the characteristics output in the step S31 through a block attention module;
step S33: and inputting the characteristics output by the block attention module into a regression network module to obtain a gas outburst prediction result.
Further, in step S31, the depth-gating convolution network module is formed by stacking multiple layers of gating convolutions, where the gating convolutions operate as follows:
in the formula, h i Conv, the output feature of the gating convolution for layer i 1 And Conv 2 Sigma is two convolution operations 1 To activate a function Sigmoid, sigma 2 For the Selu activation function,is Hadamard product;
the channel attention module is calculated as follows:
in the formula, h ca For the output characteristics of the channel attention module, h f For the output characteristics of the depth gating convolution module, MP is the maximum pooling operation, AP is the average pooling operation, and w ca Is the weight of the full connection layer, sigma ca The function is activated for Softmax.
Further, the calculation process in step S32 is as follows:
in the formula, h b As a feature of the block attention module output,for the feature of the ith branch output, concat is a join operation, w b Is the weight of the full connection layer, sigma b The function is activated for Softmax.
Further, the calculation process in step S33 is as follows:
in the method, in the process of the invention,for the prediction result of the network, w y Is all thatWeights, sigma, of the connection layer y The function is activated for Sigmoid.
Further, in step S4, a loss function and Adam algorithm are adopted to perform network training, where the loss function is a mean square error loss function.
Further, in step S5, a threshold value for gas outburst is set, and the gas outburst data after data preprocessing is input into a depth prediction model to predict the occurrence probability of gas outburst;
if the probability of the occurrence of the predicted gas outburst is greater than the threshold value, the occurrence of the gas outburst is indicated; otherwise, it indicates that no gas outburst occurred.
The invention has the beneficial effects that:
according to the deep learning method, the gas salient data collected by the three sensors are used as the input of a deep prediction network, and the multi-branch deep gating convolution network module is used as the feature extraction module, so that the hidden features of multi-source historical data are fully mined. The extracted depth features are screened and fused by adopting different attention mechanisms, so that the complexity of the features can be reduced, the data features can be deeply mined, and the prediction accuracy can be improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the description of the embodiments or the prior art will be briefly described, and it is obvious that the drawings in the description below are some embodiments of the present invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic illustration of the process of the present invention;
FIG. 2 is a schematic diagram of a depth-gated convolutional network module of the present invention;
FIG. 3 is a schematic diagram of a channel attention module of the present invention;
FIG. 4 is a block attention module schematic of the present invention;
fig. 5 is a schematic diagram of a regression network module of the present invention.
Detailed Description
The following description of the technical solution in the embodiments of the present invention is clear and complete. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways other than those described herein, and persons skilled in the art will readily appreciate that the present invention is not limited to the specific embodiments disclosed below.
According to the coal mine gas outburst prediction deep learning method based on multi-mode data fusion, namely, multi-branch and attention mechanisms are adopted in a deep convolution network, multi-source data information is fully utilized, and the robustness of the method and the accuracy of gas outburst prediction are improved.
The invention aims to provide a coal mine gas outburst prediction deep learning method based on multi-mode data fusion, which comprises the steps of acquiring acoustic emission data, gas pressure data and coal resistivity data by using a sensor; interpolation is carried out on missing values existing in the data by utilizing a KNN algorithm, and then normalization processing is carried out on the data; constructing a multi-branch depth gating convolution network based on an attention mechanism; adopting the mean square error as a loss function of the network, and optimizing network parameters by utilizing Adam; and predicting the probability of occurrence of gas outburst by using the trained network model to obtain a final result.
In connection with fig. 1, a schematic diagram of the deep learning method of the present invention: the invention provides a coal mine gas outburst prediction deep learning method based on multi-mode data fusion, which comprises the following specific steps:
step S1: acquiring gas outburst data under a plurality of different gas states by using different sensors; the gas outburst data comprise acoustic emission data X 1 Data X of gas pressure 2 And coal resistivity data X 3 The three data are three modality data, which are obtained by 3 different sensors.
The three types of data related to the gas acquired by the three sensors at the same time are called gas outburst data, and the three types of data have relevance in time space. Acoustic emission data X 1 The method is characterized in thatIncluding amplitude, energy, signal bandwidth, peak frequency, and peak frequency amplitude; gas pressure data X 2 The method is characterized by comprising a change amplitude and a climbing event; coal resistivity data X 3 The characteristics include kurtosis coefficient, skewness coefficient and inflection point resistivity.
Step S2: carrying out data preprocessing on the gas outburst data;
the data preprocessing is to transmit acoustic emission data X of the gas outburst data 1 Data X of gas pressure 2 And coal resistivity data X 3 The missing value processing and the normalization processing are performed, respectively, as shown in fig. 1.
The missing value processing is to delete the data with more missing values directly, and the data with less missing values are subjected to missing value interpolation through a KNN algorithm. The missing value interpolation is not limited to the KNN algorithm, and other algorithms that can realize the missing value interpolation may be used.
The core idea of the KNN algorithm is to determine K samples that are spatially similar or close to the training set by distance measurement, where the training set refers to a series of data collected by a sensor and the samples refer to a data under the training set. The K samples are then used to estimate the value of the missing data. For example c x Obtaining K points { c } adjacent to a missing value of a certain point in data through Euclidean distance 1 ,c 2 ,Λ,c K Then the estimate for cx is:
wherein d i Is the distance between the missing value and its i-th neighbor.
And respectively carrying out normalization processing on the gas projection data after the missing value interpolation, as shown in fig. 1. The normalization processing is to restrict the gas salient data after the interpolation of the missing values to the range of 0,1 by adopting normalization,
wherein x is min For data minimum in training set, x max For the maximum value of the data in the training set,for the normalized values, x is the raw data value (the gas projection data before pretreatment).
Step S3: and (3) constructing a depth prediction network for coal mine gas outburst prediction, and inputting gas outburst data after the data pretreatment in the step (S2).
The depth prediction network comprises a feature extraction module, a block attention module and a regression network module which are connected in sequence; the feature extraction module includes three branches, each including a depth-gated convolutional network module and a channel attention module. Acoustic emission data X of the gas outburst data are processed correspondingly by the three branches respectively 1 Data X of gas pressure 2 And coal resistivity data X 3 . The acoustic emission data X after normalization processing in the step S2 1 Data X of gas pressure 2 And coal resistivity data X 3 The characteristics are respectively input into the constructed depth prediction network and are respectively extracted by the characteristic extraction module, the block attention module is used for fusing the characteristics of the three branches of the characteristic extraction module, and the regression network module obtains the gas outburst prediction result.
The specific input process is as follows:
step S31: acoustic emission data X of normalized gas outburst data 1 Data X of gas pressure 2 And coal resistivity data X 3 And respectively inputting the depth gating convolutional network module and the channel attention module of the corresponding branch to extract the characteristics.
The depth-gated convolutional network module is formed by stacking multiple layers of gated convolutions, as shown in FIG. 2, the structure operates faster than LSTM and GRU and has stronger fitting capacity than a structure using only normal convolutions. The depth-gated convolutional network module operates as follows:
in the formula, h i Conv, the output feature of the gating convolution for layer i 1 And Conv 2 Sigma is two convolution operations 1 To activate a function Sigmoid, sigma 2 For the Selu activation function,is Hadamard product; in the gated convolution, selu is used as the activation function.
After the depth-gating convolutional network module extracts the features, the channel attention module continues to extract the features, as shown in fig. 3, and the specific calculation process is as follows:
the channel attention module for each branch is calculated as follows:
in the formula, h ca For the output characteristics of the channel attention module, h f For the output characteristics of the depth gating convolution module, MP is the maximum pooling operation, AP is the average pooling operation, and w ca Is the weight of the full connection layer, sigma ca The function is activated for Softmax.
The channel attention module specifically operates as follows: will last layer of features h f Respectively carrying out maximum pooling operation MP and average pooling operation AP to obtain a feature 1 and a feature 2, adding corresponding positions of the feature 1 and the feature 2 to obtain a feature 3, and sequentially passing the feature 3 through a full connection layer w ca And Softmax activation function sigma ca Get and input feature h f Vector with the same dimension of the channel, and finally multiplying the vector with the channel corresponding to the input characteristic to obtain the output characteristic h ca 。
After completion, the channel attention module of each branch outputs a characteristic respectively;
step S32: the features output by the channel attention modules of the three branches are all input into the block attention module for feature fusion, as shown in fig. 4, and the specific calculation process is as follows:
in the formula, h b As a feature of the block attention module output,for the feature of the ith branch output, concat is a join operation, w b Is the weight of the full connection layer, sigma b Activating a function for Softmax; MP is the max pooling operation and AP is the average pooling operation.
Step S33: the characteristics output by the block attention module are input to the regression network module. The regression network module can obtain the gas outburst prediction result. The regression network module is obtained by connecting the full connection layer and the Sigmoid activation function in series. As shown in fig. 5, the specific calculation process is as follows:
in the method, in the process of the invention,as the gas outburst prediction result, w y Is the weight of the full connection layer, sigma y The function is activated for Sigmoid.
Step S4: training iteration of a depth prediction network for coal mine gas outburst prediction is executed, and a depth prediction model is obtained after training is completed;
and performing network training by adopting a loss function and an Adam algorithm, wherein the loss function is a mean square error loss function, and the method comprises the following steps of:
wherein M is the number of samples;is a model predictive value; y is the gas highlighting real label.
In the training process, the gas salient real labels are input into the loss function to perform back propagation training on the depth prediction network. The mean square error loss function is used as the loss of the depth prediction network, and the Adam algorithm is utilized to iteratively update the network parameters of the depth prediction network. And after the deep prediction network training is completed, obtaining a deep prediction model. The gas highlighting real label means whether gas actually occurs or not.
Step S5: and predicting the probability of occurrence of coal mine gas outburst by using the depth prediction model, and obtaining a gas outburst prediction result. The gas outburst prediction result refers to gas outburst occurrence or gas outburst non-occurrence. In the step, a threshold value of gas outburst is set, and the gas outburst data after data preprocessing is input into a depth prediction model to predict the probability of gas outburst occurrence so as to obtain a gas outburst prediction result.
If the probability of the occurrence of the predicted gas outburst is greater than the threshold value, the occurrence of the gas outburst is indicated; otherwise, it indicates that no gas outburst occurred.
It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Claims (7)
1. The deep learning method for coal mine gas outburst prediction based on multi-mode data fusion is characterized by comprising the following steps:
step S1: different sensors acquire a plurality of gas outburst data under different gas states; the gas outburst data comprise acoustic emission data, gas pressure data and coal resistivity data;
step S2: carrying out data preprocessing on the gas outburst data;
step S3: a depth prediction network for coal mine gas outburst prediction is constructed, and gas outburst data after data preprocessing is input;
step S4: training iteration of a depth prediction network for coal mine gas outburst prediction is executed to form a depth prediction model;
step S5: predicting the probability of occurrence of coal mine gas outburst by using a depth prediction model to obtain a gas outburst prediction result;
step S2 includes the steps of:
step S21: respectively carrying out missing value processing on acoustic emission data, gas pressure data and coal resistivity data of the gas outburst data;
step S22: respectively carrying out normalization processing on the acoustic emission data, the gas pressure data and the coal resistivity data of the gas outburst data subjected to the missing value processing;
in the step S21, the missing value processing is to delete the gas outburst data with more missing values, and the gas outburst data with less missing values is subjected to missing value interpolation through a KNN algorithm;
in the step S22, the normalization processing is to restrict the gas salient data after the missing value interpolation to the range of [0,1] by adopting normalization;
the depth prediction network comprises a feature extraction module, a block attention module and a regression network module which are connected in sequence; the characteristic extraction module comprises three branches, and the three branches respectively correspond to acoustic emission data, gas pressure data and coal resistivity data for processing the gas outburst data; each branch comprises a depth gating convolutional network module and a channel attention module; the block attention module fuses the features output by the three branches of the feature extraction module.
2. The deep learning method of coal mine gas outburst prediction based on multi-mode data fusion according to claim 1, wherein in step S3, the data preprocessing post-gas outburst data input process is as follows:
step S31: respectively inputting acoustic emission data, gas pressure data and coal resistivity data of the gas outburst data after data preprocessing into a depth gating convolution network module and a channel attention module of corresponding branches to perform feature extraction;
step S32: fusing the characteristics output in the step S31 through a block attention module;
step S33: and inputting the characteristics output by the block attention module into a regression network module to obtain a gas outburst prediction result.
3. The method for deep learning of coal mine gas outburst prediction based on multi-modal data fusion according to claim 2, wherein in step S31, the deep gating convolution network module is formed by stacking multi-layer gating convolutions, and the gating convolution operation is as follows:
in the formula, h i Conv, the output feature of the gating convolution for layer i 1 And Conv 2 Sigma is two convolution operations 1 To activate a function Sigmoid, sigma 2 For the Selu activation function,is Hadamard product;
the channel attention module is calculated as follows:
in the formula, h ca For the output characteristics of the channel attention module, h f For the output characteristics of the depth gating convolution module, MP is the maximum pooling operation, AP is the average pooling operation, and w ca Is the weight of the full connection layer, sigma ca The function is activated for Softmax.
4. The deep learning method of coal mine gas outburst prediction based on multi-modal data fusion according to claim 3, wherein the calculation process in step S32 is as follows:
in the formula, h b As a feature of the block attention module output,for the feature of the ith branch output, concat is a join operation, w b Is the weight of the full connection layer, sigma b The function is activated for Softmax.
5. The deep learning method of coal mine gas outburst prediction based on multi-modal data fusion according to claim 4, wherein the calculation process in step S33 is as follows:
in the method, in the process of the invention,as the gas outburst prediction result, w y Is the weight of the full connection layer, sigma y The function is activated for Sigmoid.
6. The multi-modal data fusion-based coal mine gas outburst prediction deep learning method according to claim 5, characterized in that in step S4, network training is performed by adopting a loss function and Adam algorithm, wherein the loss function is a mean square error loss function.
7. The method for deep learning coal mine gas outburst prediction based on multi-mode data fusion according to any one of claims 1 to 6, wherein in step S5, a threshold value for gas outburst is set, and the data-preprocessed gas outburst data is input into a deep prediction model to predict the occurrence probability of gas outburst;
if the probability of the occurrence of the predicted gas outburst is greater than the threshold value, the occurrence of the gas outburst is indicated; otherwise, it indicates that no gas outburst occurred.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310790338.0A CN116975775B (en) | 2023-06-29 | 2023-06-29 | Deep learning method for coal mine gas outburst prediction based on multi-mode data fusion |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310790338.0A CN116975775B (en) | 2023-06-29 | 2023-06-29 | Deep learning method for coal mine gas outburst prediction based on multi-mode data fusion |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116975775A CN116975775A (en) | 2023-10-31 |
CN116975775B true CN116975775B (en) | 2024-01-30 |
Family
ID=88475932
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310790338.0A Active CN116975775B (en) | 2023-06-29 | 2023-06-29 | Deep learning method for coal mine gas outburst prediction based on multi-mode data fusion |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116975775B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117953313B (en) * | 2024-03-26 | 2024-06-18 | 贵州大学 | Method and system for realizing anomaly identification of mine data based on artificial intelligence |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107605536A (en) * | 2017-11-02 | 2018-01-19 | 湖南科技大学 | Coal and gas prominent real-time early warning device and method based on Multi-source Information Fusion |
CN109356655A (en) * | 2018-12-06 | 2019-02-19 | 西安科技大学 | The compound coal rock dynamic disaster prediction and warning method of index is extracted based on gas density and sound emission |
CN109441547A (en) * | 2018-12-29 | 2019-03-08 | 煤炭科学技术研究院有限公司 | A kind of getting working face coal and gas prominent real-time system for monitoring and pre-warning and method |
CN111126658A (en) * | 2019-11-15 | 2020-05-08 | 太原理工大学 | Coal mine gas prediction method based on deep learning |
CN112183901A (en) * | 2020-11-06 | 2021-01-05 | 贵州工程应用技术学院 | Coal and gas outburst strength prediction method based on deep learning |
CN112308306A (en) * | 2020-10-27 | 2021-02-02 | 贵州工程应用技术学院 | Multi-mode input coal and gas outburst risk prediction method |
CN113298165A (en) * | 2021-05-31 | 2021-08-24 | 黄炳僖 | Coal mine gas disaster advanced prediction method based on multi-source information fusion |
-
2023
- 2023-06-29 CN CN202310790338.0A patent/CN116975775B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107605536A (en) * | 2017-11-02 | 2018-01-19 | 湖南科技大学 | Coal and gas prominent real-time early warning device and method based on Multi-source Information Fusion |
CN109356655A (en) * | 2018-12-06 | 2019-02-19 | 西安科技大学 | The compound coal rock dynamic disaster prediction and warning method of index is extracted based on gas density and sound emission |
CN109441547A (en) * | 2018-12-29 | 2019-03-08 | 煤炭科学技术研究院有限公司 | A kind of getting working face coal and gas prominent real-time system for monitoring and pre-warning and method |
CN111126658A (en) * | 2019-11-15 | 2020-05-08 | 太原理工大学 | Coal mine gas prediction method based on deep learning |
CN112308306A (en) * | 2020-10-27 | 2021-02-02 | 贵州工程应用技术学院 | Multi-mode input coal and gas outburst risk prediction method |
CN112183901A (en) * | 2020-11-06 | 2021-01-05 | 贵州工程应用技术学院 | Coal and gas outburst strength prediction method based on deep learning |
CN113298165A (en) * | 2021-05-31 | 2021-08-24 | 黄炳僖 | Coal mine gas disaster advanced prediction method based on multi-source information fusion |
Non-Patent Citations (2)
Title |
---|
基于残缺数据的煤与瓦斯突出预测系统;汝彦冬 等;《哈尔滨商业大学学报(自然科学版)》;第37卷(第2期);摘要,第1.3节 * |
梁北矿煤与瓦斯突出危险性监测技术研究;张酉年 等;煤炭技术;第37卷(第11期);摘要,第2-3.1节 * |
Also Published As
Publication number | Publication date |
---|---|
CN116975775A (en) | 2023-10-31 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN116975775B (en) | Deep learning method for coal mine gas outburst prediction based on multi-mode data fusion | |
CN111967486A (en) | Complex equipment fault diagnosis method based on multi-sensor fusion | |
CN113723010B (en) | Bridge damage early warning method based on LSTM temperature-displacement correlation model | |
CN110443488A (en) | The recognition methods of drilling well spill hazard, system and equipment based on convolutional neural networks | |
CN102340811A (en) | Method for carrying out fault diagnosis on wireless sensor networks | |
CN115618296A (en) | Dam monitoring time sequence data anomaly detection method based on graph attention network | |
CN108595803A (en) | Shale gas well liquid loading pressure prediction method based on recurrent neural network | |
CN115169724A (en) | Runoff prediction method based on space-time graph convolutional neural network | |
CN116484740A (en) | Line parameter identification method based on space topology characteristics of excavated power grid | |
CN117726181B (en) | Collaborative fusion and hierarchical prediction method for typical disaster risk heterogeneous information of coal mine | |
CN115953900A (en) | Traffic flow prediction method based on multi-dimensional time and space dependence mining | |
CN114897007B (en) | Drilling machine health condition assessment method of composite information hierarchical convolutional neural network | |
CN115827335A (en) | Time sequence data missing interpolation system and method based on modal crossing method | |
CN113642235B (en) | Icing prediction method based on local space-time feature fusion | |
CN116894180B (en) | Product manufacturing quality prediction method based on different composition attention network | |
CN117556310A (en) | Spacecraft residual life prediction method | |
CN117768235A (en) | Real-time flow monitoring alarm system based on Internet of things | |
CN112528548A (en) | Self-adaptive depth coupling convolution self-coding multi-mode data fusion method | |
CN112215197A (en) | Underground cable fault early warning method and system based on artificial intelligence | |
CN115859090A (en) | Turnout fault diagnosis method and system based on 1DCNN-LSTM | |
CN114970336A (en) | Traffic flow prediction method based on improved GCN-attention algorithm | |
CN113761795A (en) | Aircraft engine fault detection method and system | |
CN110348639A (en) | A kind of coal mine roof plate gushing water danger classes prediction technique | |
CN117493837B (en) | Machine learning-based shield tunneling machine attitude item prediction method | |
CN117171528B (en) | High-altitude overhanging structure health monitoring method based on deep convolution fusion memory network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |