A ventilation data analysis method and system for intelligent ventilation in a mine

Abstract

The present invention belongs to the technical field of mine ventilation data analysis, and specifically provides a ventilation data analysis method and system for intelligent mine ventilation. The method mainly includes: obtaining mine environment parameters, where the mine environment parameters include gas concentration, real-time air volume, and temperature; calculating a dynamic risk index based on the mine environment parameters; generating a risk trend status code according to the dynamic risk index; obtaining an atmospheric pressure sequence within a specified past time range, and processing the atmospheric pressure sequence using an adaptive atmospheric pressure model. The adaptive atmospheric pressure model is based on a temporal convolutional network and dynamically determines the optimal dilation rate of the dilated convolutional layer. This disclosure quantifies the environmental evolution law based on real-time risk trend coding, combines multi-scale pressure features adaptively extracted, and at the same time, by integrating the dynamic association between the tunneling state and the risk trend, can effectively decouple the coupling effect of mechanical disturbance and pressure fluctuation, early warn of abnormal fluctuations in gas concentration, and significantly improve the real-time safety prevention and control under complex working conditions.

Description

Technical Field

[0001] The present invention belongs to the technical field of mine ventilation data analysis, and particularly relates to a ventilation data analysis method and system for intelligent mine ventilation. Background Art

[0002] When mining a mine, harmful gases such as gas are generated, and mechanical operations will bring dust. The narrow space is prone to sultry and oxygen-deficient air. Mine ventilation can discharge harmful gases and dust, supplement fresh air, reduce the temperature, create a safe and comfortable working environment for miners, prevent dangers such as gas accumulation and explosion, and ensure the smooth progress of production.

[0003] The dynamic imbalance of the mine ventilation system stems from the multi-source coupling of gas, air pressure, and mechanical disturbances. Nonlinear time-varying correlations cause ventilation predictions to continuously deviate from the true demand, inducing local gas overrun and a sharp increase in energy consumption. Traditional methods rely on static models and fixed-time window analysis of air pressure data, masking the interaction laws between short-term mutations and long-term trends, and it is difficult to capture the dynamic correlations of multi-scale fluctuations and risk evolution of air pressure. Therefore, it is inconvenient to analyze the superimposed effects of mechanical disturbances on time series characteristics, and it is difficult to correct the parameter coupling deviation under complex working conditions, which easily leads to safety prevention and control lagging behind risk evolution. Summary of the Invention

[0004] The present disclosure effectively solves the problem that the prior art relies on static models and fixed-time window analysis of air pressure data, which easily leads to safety prevention and control lagging behind risk evolution, by providing a ventilation data analysis method and system for intelligent mine ventilation, and can early warn of abnormal fluctuations in gas concentration, significantly improving the real-time performance of safety prevention and control under complex working conditions.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] In a first aspect, the present disclosure provides a ventilation data analysis method for intelligent mine ventilation, including: obtaining mine tunnel environment parameters, where the mine tunnel environment parameters include gas concentration, real-time air volume, and temperature; calculating a dynamic risk index according to the mine tunnel environment parameters; generating a risk trend status code according to the dynamic risk index; obtaining an atmospheric pressure sequence within a specified past time range, and processing the atmospheric pressure sequence by using an adaptive atmospheric pressure model, where the adaptive atmospheric pressure model is based on a time convolutional network, dynamically determines the optimal dilation rate of the dilated convolutional layer, and outputs a time series feature vector after integrating multi-scale information in depth; obtaining the tunneling state of a roadheader; and obtaining a ventilation volume prediction interval for a specified future time range according to the time series feature vector, the risk trend status code, the tunneling state of the roadheader, and the real-time air volume.

[0007] Further, obtain the mine tunnel environment parameters, including: configuring basic acquisition parameters, including: setting the sampling frequencies of gas concentration, real-time air volume, and temperature, defining the time window and the space window, as well as the sliding intervals of the time window and the space window; based on the basic acquisition parameters, generate a three-dimensional time series tensor including a time axis, a spatial node distribution, and a parameter channel according to the sliding of the time window; dynamically crop the three-dimensional time series tensor according to the sliding of the space window to obtain effective node data.

[0008] Further, defining the space window includes: taking the roadheader as a reference point, using an ultra-wideband positioning system to obtain the real-time coordinates of the reference point; determining the coordinates of the sensor, calculating the actual distance between the sensor and the reference point; screening out the sensors whose actual distances meet the preset distance.

[0009] Further, calculate the dynamic risk index according to the mine tunnel environment parameters, including: for the effective node data within the current time window and space window: calculate the mean value of the gas concentration according to the sampling frequency and spatial node distribution of the gas concentration; normalize the real-time air volume, and calculate the mean value of the normalized real-time air volume according to the sampling frequency and spatial node distribution of the real-time air volume; calculate the mean value of the temperature according to the sampling frequency and spatial node distribution of the temperature; calculate the gas concentration risk factor, the air volume risk factor, and the temperature risk factor according to the mean value of the gas concentration, the mean value of the normalized real-time air volume, and the mean value of the temperature; calculate the dynamic risk index according to the gas concentration risk factor, the air volume risk factor, and the temperature risk factor.

[0010] Further, generate a risk trend status code according to the dynamic risk index, including: determining two or more consecutive increasing risk levels, each risk level corresponding to a unique code; defining boundary thresholds that are one less than the number of risk levels, and arranging all boundary thresholds in ascending order; determining the boundary threshold interval where the dynamic risk index is located, and outputting the risk level code according to the code mapping mechanism.

[0011] Among them, the code mapping mechanism includes: mapping the dynamic risk index lower than the minimum boundary threshold to the lowest risk level code; mapping the dynamic risk index between every two adjacent boundary thresholds to the intermediate risk level codes in the order of the boundary threshold arrangement; mapping the dynamic risk index higher than or equal to the maximum boundary threshold to the highest risk level code.

[0012] Further, the adaptive atmospheric pressure model includes an input layer, a first specified number of dilated convolutional layers, and an output layer; the input layer is used to receive the atmospheric pressure sequence within a specified past time range; the first specified number of dilated convolutional layers are assigned optimal dilation rates to capture the variation characteristics of the atmospheric pressure at different periods, obtaining multiple feature maps, concatenating the feature maps by channels, and integrating multi-scale information into a feature tensor; the output layer flattens the feature tensor to obtain a time series feature vector.

[0013] Further, in the process of assigning optimal dilation rates to the first specified number of dilated convolutional layers, obtaining the optimal dilation rate includes: acquiring the atmospheric pressure sequence within a specified past time range; performing multi-scale decomposition on the atmospheric pressure sequence to obtain multiple subsequences; extracting the statistical features of the subsequences to obtain feature vectors; clustering the feature vectors of all subsequences to obtain a second specified number of cluster centers and the cluster labels to which each subsequence belongs; classifying initial dilation rates for each cluster center to obtain a set of initial dilation rates; when training the adaptive atmospheric pressure model, calculating the importance weights of the subsequences corresponding to each cluster in the current input; fine-tuning the initial dilation rates according to the importance weights of the subsequences corresponding to each cluster in the current input to obtain the final dilation rates; constructing a set of final dilation rates, using the set of final dilation rates to train the adaptive atmospheric pressure model and feedback adjustment to obtain the optimal dilation rate.

[0014] Further, according to the time series feature vector, the risk trend state encoding, the tunneling state of the tunneling machine, and the real-time air volume, obtaining the ventilation volume prediction interval for a specified future time range includes: using the GBRT improved architecture to process the time series feature vector, the risk trend state encoding, the tunneling state of the tunneling machine, and the real-time air volume, and outputting the ventilation volume prediction interval for a specified future time range.

[0015] Among them, the GBRT improved architecture includes: an input layer, a feature fusion layer, a gradient boosting tree ensemble layer, and an output layer; the input layer is used to receive the time series feature vector, the risk trend state encoding, the tunneling state of the tunneling machine, and the real-time air volume; the feature fusion layer is used to concatenate the time series feature vector and the risk trend state encoding vector by channels to form a fused feature; performing a sliding window mean filter on the real-time air volume to generate an air volume trend feature; concatenating the fused feature, the air volume trend feature, and the tunneling state parameters into a joint feature vector; the gradient boosting tree ensemble layer contains a third specified number of regression trees, and each regression tree performs node splitting based on the joint feature vector to output the ventilation volume prediction base value for each time step; the output layer is used to perform quantile regression on the ventilation volume prediction base value, calculate the upper bound quantile and the lower bound quantile for a specified future time range using a quantile loss function, and generate the ventilation volume prediction interval.

[0016] Further, the quantile output layer dynamically scales the upper and lower bounds of the ventilation volume prediction interval according to the ratio of the current dynamic risk index to the historical risk interval, and the scaling coefficient is positively correlated with the dynamic risk index.

[0017] In a second aspect, the present disclosure provides a ventilation data analysis system for intelligent mine ventilation, which includes: a multi-source environmental parameter perception module, a dynamic risk assessment and fusion module, a risk coding and barometric pressure time series feature generation module, and a multi-modal ventilation decision module.

[0018] The multi-source environmental parameter perception module is used to obtain mine tunnel environmental parameters, and the mine tunnel environmental parameters include gas concentration, real-time air volume, and temperature; the dynamic risk assessment and fusion module is used to calculate the dynamic risk index according to the mine tunnel environmental parameters; the risk coding and barometric pressure time series feature generation module is used to generate a risk trend state code according to the dynamic risk index; obtain the barometric pressure sequence within a specified past time range, and process the barometric pressure sequence using an adaptive barometric pressure model. The adaptive barometric pressure model is based on a temporal convolutional network, dynamically determines the optimal dilation rate of the dilated convolutional layer, and outputs a time series feature vector after deep integration of multi-scale information; the multi-modal ventilation decision module is used to obtain the tunneling state of the roadheader; according to the time series feature vector, the risk trend state code, the tunneling state of the roadheader, and the real-time air volume, obtain the ventilation volume prediction interval for a specified future time range.

[0019] In a third aspect, the present disclosure provides a device, which includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement the steps of the ventilation data analysis method for intelligent mine ventilation as described in the first aspect when executing the computer program.

[0020] In a fourth aspect, the present disclosure provides a storage medium, in which computer program instructions are stored. When the computer program instructions are read and run by a processor, the steps of the ventilation data analysis method for intelligent mine ventilation as described in the first aspect are executed.

[0021] Advantages of the present invention:

[0022] Based on real-time risk trend coding to quantify the environmental evolution law, combined with adaptively extracted multi-scale barometric pressure features, and at the same time by integrating the dynamic association between the tunneling state and the risk trend, the present disclosure effectively solves the problem that the prior art relies on static models and fixed time windows to analyze barometric pressure data, which easily leads to the lag of safety prevention and control behind the risk evolution. It can effectively decouple the coupling effect of mechanical disturbance and barometric pressure fluctuation, early warn of abnormal fluctuations in gas concentration, and significantly improve the real-time performance of safety prevention and control under complex working conditions.

[0023] Other features and advantages of the present invention will be set forth in the following description, and in part will be obvious from the description, or may be learned by practice of the present invention. The objectives and other advantages of the present invention can be realized and attained by the structure particularly pointed out in the specification and the drawings. Description of the Drawings

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0025] Figure 1 The flowchart shows a method for analyzing ventilation data of intelligent ventilation in a mine according to the present invention;

[0026] Figure 2 The block diagram shows a system for analyzing ventilation data of intelligent ventilation in a mine according to the present invention. Detailed Embodiments

[0027] To solve the problems proposed in the background art, the present disclosure quantifies the environmental evolution law based on real-time risk trend coding, combines multi-scale air pressure characteristics extracted adaptively, and at the same time, by integrating the dynamic association between the tunneling state and the risk trend, early warns of abnormal fluctuations in gas concentration, significantly improving the real-time safety prevention and control under complex working conditions.

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0029] In some embodiments, as Figure 1 shown, the present disclosure provides a method for analyzing ventilation data of intelligent ventilation in a mine, including:

[0030] S100. Obtain the mine environment parameters, which include gas concentration, real-time air volume, and temperature.

[0031] S200. Calculate the dynamic risk index according to the mine environment parameters.

[0032] S300. Generate a risk trend status code according to the dynamic risk index; obtain an atmospheric pressure sequence within a specified past time range, and process the atmospheric pressure sequence using an adaptive atmospheric pressure model. The adaptive atmospheric pressure model is based on a temporal convolutional network, dynamically determines the optimal dilation rate of the dilated convolutional layer, and outputs a temporal feature vector after integrating multi-scale information in depth.

[0033] S400. Obtain the tunneling state of the tunneling machine; obtain a predicted ventilation volume interval for a specified future time range based on the temporal feature vector, the risk trend status code, the tunneling state of the tunneling machine, and the real-time air volume.

[0034] In some embodiments, obtaining the mine environment parameters in S100 includes:

[0035] S110. Configure basic acquisition parameters, including: setting the sampling frequencies of gas concentration, real-time air volume, and temperature, defining a time window and a space window, as well as the sliding intervals of the time window and the space window.

[0036] Set the sampling frequencies of gas concentration, real-time air volume, and temperature. For example: the sampling frequency of gas concentration is 1 time per second, the sampling frequency of real-time air volume is 2 times per second, and the sampling frequency of temperature is 1 time per second.

[0037] Define the sizes and sliding intervals of the time window and the space window:

[0038] For example, set the time window length to 10 minutes and the sliding interval of the time window to 5 minutes, that is, the window slides forward every 5 minutes, and the window contains 10 minutes of continuous data.

[0039] For example, with the current position of the tunneling machine as the center, define the space window as a circular area with a radius of 5 meters, and the sliding interval of the space window is 1 meter, that is, when the tunneling machine moves 1 meter, the space window is repositioned.

[0040] S120. Based on the basic acquisition parameters, generate a three-dimensional temporal tensor including a time axis, a spatial node distribution, and a parameter channel according to the sliding of the time window.

[0041] Time axis construction: Generate a time step sequence according to the sampling frequency within each time window. For example, a 10-minute time window contains 600 time steps.

[0042] Spatial node distribution: Within the space window, obtain the coordinates of each sensor through an ultra-wideband positioning system to generate a spatial node distribution matrix. For example, if there are 5 sensors in the current space window, the number of spatial nodes is 5.

[0043] Each spatial node collects three parameters: gas concentration, real-time air volume, and temperature at each time step, forming the parameter channel dimension.

[0044] The finally generated three-dimensional time-series tensor structure is: tensor shape = (number of time steps, number of spatial nodes, 3).

[0045] S130. Dynamically crop the three-dimensional time-series tensor according to the sliding of the spatial window to obtain effective node data.

[0046] S131. Obtain the real-time coordinates of the roadheader according to the ultra-wideband positioning system, and calculate the Euclidean distance between each sensor and the roadheader : ; where, ( , , ) represents the coordinates of the roadheader, and ( , , ) represents the coordinates of the i-th sensor.

[0047] S132. Screen out the sensor nodes with Euclidean distance ≤ 5 meters to generate a list of effective node indices.

[0048] S133. Extract the data of the corresponding spatial nodes from the original three-dimensional time-series tensor according to the list of effective node indices to obtain effective node data.

[0049] For example, the shape of the original three-dimensional time-series tensor is 600×20×3. After filtering by the spatial window, 12 effective nodes are retained, and the shape of the cropped effective node data is 600×12×3.

[0050] In some embodiments, defining the spatial window includes:

[0051] S111. Taking the roadheader as the reference point, use the ultra-wideband positioning system to obtain the real-time coordinates of the reference point.

[0052] A UWB positioning tag can be installed on the roadheader body, and the tag refresh frequency is 10 times per second. The UWB base station receives the tag signal and calculates the real-time coordinates of the roadheader ( , , ) through the time difference of arrival algorithm.

[0053] S112. Determine the coordinates of the sensor and calculate the actual distance between the sensor and the reference point.

[0054] All sensors (gas, air volume, temperature sensors) in the mine tunnel are pre-set with UWB positioning tags, and their coordinates ( , , ) are obtained through the same UWB system.

[0055] S113. Screen out the sensors whose actual distance meets the preset distance.

[0056] Calculate the Euclidean distance between each sensor and the reference point of the roadheader. If the defined spatial window radius is 5 meters, the screening condition is the Euclidean distance ≤ 5 meters.

[0057] Calculate the Euclidean distance between all sensors and the roadheader in real time , and generate a list of valid sensor indices. For example, there are 20 sensors in the mine tunnel, and 12 in the current window meet the ≤ 5 meters, then these 12 sensors are selected.

[0058] The data of the selected sensors directly reflects the state of high-risk areas around the roadheader, which can improve the accuracy of subsequent risk index calculation.

[0059] In some embodiments, calculate the dynamic risk index according to the mine tunnel environmental parameters, including:

[0060] 210. For the valid node data within the current time window and spatial window: Calculate the mean value of the gas concentration according to the sampling frequency and spatial node distribution of the gas concentration; Normalize the real-time air volume, and calculate the mean value of the normalized real-time air volume according to the sampling frequency and spatial node distribution of the real-time air volume; Calculate the mean value of the temperature according to the sampling frequency and spatial node distribution of the temperature.

[0061] Take the arithmetic mean of the gas concentration values of all sensors in the valid node data to obtain the mean value of the gas concentration .

[0062] Scale the real-time air volume value to [0, 1], and the formula can be used: ; where and represent the minimum and maximum values of the real-time air volume of all sensors within the current spatial window respectively, represents the i-th real-time air volume value before scaling, represents the i-th real-time air volume value after scaling.

[0063] Take the arithmetic mean of the air volume values of all sensors in the valid node data to obtain the mean value of the air volume .

[0064] Take the arithmetic mean of the temperature values of all sensors in the valid node data to obtain the mean value of the temperature .

[0065] 220. Calculate the gas concentration risk factor, air volume risk factor and temperature risk factor according to the mean value of the gas concentration, the mean value of the normalized real-time air volume and the mean value of the temperature.

[0066] Calculate the gas concentration risk factor based on the deviation degree between the average gas concentration and the safety threshold: ; where represents the gas concentration risk factor, represents the gas concentration safety threshold.

[0067] Calculate the air volume risk factor based on the deviation between the normalized average air volume and the ideal value: ; where represents the air volume risk factor.

[0068] Calculate the temperature risk factor based on the deviation of the average temperature from the standard operation temperature range:

[0069]

[0070] where represents the minimum value of the standard operation temperature range, such as 20°C, represents the maximum value of the standard operation temperature range, such as 35°C.

[0071] 230. Calculate the dynamic risk index based on the gas concentration risk factor, the air volume risk factor, and the temperature risk factor.

[0072] The dynamic risk index is: ; where represents the dynamic risk index, , , respectively represent the weights of the gas risk factor, the air volume risk factor, and the temperature risk factor.

[0073] By comprehensively considering the synergistic effects of gas, air volume, and temperature through the dynamic risk index, the accuracy of early warning can be improved, , , and it can be dynamically adjusted according to the tunneling stage.

[0074] For the historical data of each tunneling stage, the information entropy of the gas concentration risk factor, the air volume risk factor, and the temperature risk factor can be calculated respectively. By calculating the proportion of the information entropy of each factor in the total information entropy, the weights , , are determined.

[0075] In some embodiments, in S300, generating a risk trend status code according to the dynamic risk index includes:

[0076] Sa310. Determine two or more consecutive increasing risk levels, each risk level corresponding to a unique code; define a boundary threshold that is one less than the number of risk levels, and all boundary thresholds are arranged in ascending order.

[0077] Sa320. Determine the boundary threshold interval where the dynamic risk index is located, and output the risk level code according to the coding mapping mechanism.

[0078] Among them, the coding mapping mechanism includes:

[0079] Map the dynamic risk index lower than the minimum boundary threshold to the lowest risk level code;

[0080] Map the dynamic risk index between every two adjacent boundary thresholds to the intermediate risk level codes in the order of the boundary threshold arrangement;

[0081] Map the dynamic risk index higher than or equal to the maximum boundary threshold to the highest risk level code.

[0082] For example, define three risk levels: low risk (code 0), medium risk (code 1), high risk (code 2), and two boundary thresholds (denoted as = 0.3 and = 0.6), and all boundary thresholds are arranged in ascending order as , and the boundary thresholds are determined by historical data statistics or expert experience. For example, take the 50% quantile of the dynamic risk index in the historical accidents of the mine as , and the 80% quantile as .

[0083] Match the current dynamic risk index with the boundary threshold sequence for interval matching:

[0084] If , then it is determined as low risk and mapped to code 0.

[0085] If , then it is determined as medium risk and mapped to code 1.

[0086] If , then it is determined as high risk and mapped to code 2.

[0087] In some embodiments, the adaptive atmospheric pressure model includes an input layer, a first specified number of dilated convolutional layers, and an output layer.

[0088] The input layer is used to receive the atmospheric pressure sequence within a specified past time range.

[0089] The first specified number of dilated convolutional layers are assigned optimal dilation rates to capture the variation characteristics of the atmospheric pressure at different periods, obtain multiple feature maps, and splice the feature maps by channels to integrate multi-scale information into a feature tensor.

[0090] The output layer flattens the feature tensor to obtain a time series feature vector.

[0091] The input layer receives the atmospheric pressure sequence within a specified past time range, such as the atmospheric pressure data obtained at a sampling frequency of once per second within the past 24 hours, to form a one-dimensional input vector. The dimension of the input data is L×1, where L represents the time step.

[0092] The first specified number of dilated convolutional layers, such as 3 layers, and each layer is assigned an optimally determined dilation rate, such as the dilation rate =2, =4, =8. The dilated convolutional layer captures the atmospheric pressure change patterns at different time scales by adjusting the dilation rate.

[0093] For the k-th dilated convolution layer, its dilation rate is , the convolutional kernel size is K, and this layer performs a dilated convolution operation on the input sequence X to output the feature map , and the formula is: ; where represents the weight at the i-th position in the convolutional kernel of the k-th layer, and t is the current time step.

[0094] Different dilation rates correspond to different receptive field ranges. For example, the convolutional layer with dilation rate d = 2 captures short-period fluctuations, such as minute-level changes, while the layer with d = 8 captures long-period trends, such as hour-level changes.

[0095] The feature maps output by each dilated convolutional layer, such as , , , are concatenated along the channel dimension to form a multi-scale feature tensor , enabling the model to simultaneously retain the air pressure change information at different time granularities.

[0096] The output layer flattens the feature tensor into a one-dimensional time series feature vector V for input to the subsequent ventilation volume prediction model. The flattening operation can be achieved through a fully connected layer or global pooling.

[0097] By dynamically allocating the optimal dilation rate, the adaptive atmospheric pressure model can automatically select the dilation rate combination according to the periodic characteristics of the current input sequence.

[0098] For example, when high-frequency oscillations are detected in the air pressure, the model assigns a smaller dilation rate to the shallow convolution to capture details; when the air pressure shows a slow trend change, the deep convolution uses a large dilation rate to expand the receptive field, significantly enhancing the representation ability of the time series feature vector for complex air pressure patterns.

[0099] In some embodiments, among the first specified number of dilated convolutional layers being assigned the optimal dilation rate, obtaining the optimal dilation rate includes:

[0100] Sb310. Obtain the atmospheric pressure sequence within a specified past time range.

[0101] For example, for data with a sampling interval of 1 minute in the past 24 hours, the total length L = 1440.

[0102] Sb320. Perform multi-scale decomposition on the atmospheric pressure sequence to obtain multiple subsequences.

[0103] Perform multi-scale decomposition using wavelet transform to obtain multiple subsequences.

[0104] For example, use 3-layer discrete wavelet decomposition to generate an approximate coefficient subsequence (low-frequency trend) and detail coefficient subsequences (high-frequency noise), (medium-frequency fluctuations), (low-frequency period), a total of 4 subsequences, and the length of each subsequence is the same as the original sequence.

[0105] Sb330. Extract the statistical features of the subsequences to obtain a feature vector.

[0106] The following statistical features can be extracted for each subsequence to form a feature vector:

[0107] Time-domain features: mean, variance, skewness, kurtosis.

[0108] Frequency-domain feature: Calculate the main frequency amplitude through fast Fourier transform.

[0109] Nonlinear feature: approximate entropy.

[0110] Sb340. Cluster the feature vectors of all subsequences to obtain the second specified number of cluster centers and the cluster labels to which each subsequence belongs.

[0111] Perform K-means clustering on the feature vectors of all subsequences. The second specified number can be 3 to obtain the cluster centers , , and the cluster labels of each subsequence.

[0112] Sb350. Classify the initial inflation rate for each cluster center to obtain a set of initial inflation rates.

[0113] If the subsequence belongs to the high-frequency feature cluster, such as corresponding to , assign a smaller inflation rate d = 2 to capture details.

[0114] If it belongs to the medium-frequency cluster, such as corresponding to , assign a medium inflation rate d = 4.

[0115] If it belongs to low-frequency clustering, such as the corresponding , a relatively large equal expansion rate d = 8 is assigned.

[0116] Then the initial expansion rate set is {2, 4, 8}.

[0117] Sb360. When training the adaptive atmospheric pressure model, calculate the importance weight of the subsequence corresponding to each cluster in the current input.

[0118] Calculate the importance weight of each cluster subsequence in the current input through the attention mechanism. For the k-th cluster, the weight is .

[0119] Sb370. According to the importance weight of the subsequence corresponding to each cluster in the current input, fine-tune the initial expansion rate to obtain the final expansion rate.

[0120] According to the weight scale the initial expansion rate . For example, if = 0.6 (dominated by high-frequency features), then adjust the corresponding expansion rate to , where represents the adjustment coefficient to obtain the final expansion rate .

[0121] Sb380. Construct a set of final expansion rates, use the set of final expansion rates to train the adaptive atmospheric pressure model and perform feedback adjustment to obtain the optimal expansion rate.

[0122] Input the adjusted expansion rate set into the temporal convolutional network for forward propagation, calculate the prediction error using the MAE loss function, and update the model parameters through backpropagation.

[0123] Meanwhile, dynamically update the cluster center according to the validation set performance. For example, every 10 training epochs, re-execute Sb320 - Sb350 with the latest data to update the cluster center and the expansion rate allocation rule, forming a closed-loop optimization.

[0124] During the blasting operation of the roadheader, instantaneous spikes will appear in the atmospheric pressure sequence, resulting in a sharp increase in high-frequency components. At this time, by dynamically increasing the weight of the high-frequency cluster subsequence , and reducing its expansion rate to d = 1, so that the model convolution kernel can cover a finer time granularity, such as second-level changes, to accurately capture the characteristics of sudden air pressure fluctuations.

[0125] In some embodiments, according to the temporal feature vector, the risk trend state encoding, the driving state of the roadheader, and the real-time air volume, obtain the ventilation volume prediction interval for a future specified time range, including:

[0126] The GBRT improved architecture is adopted to process the time series feature vector, risk trend status encoding, tunneling state of the roadheader, and real-time air volume, and output the ventilation volume prediction interval for a specified future time range.

[0127] Among them, the GBRT improved architecture includes: an input layer, a feature fusion layer, a gradient boosting tree ensemble layer, and an output layer.

[0128] The input layer is used to receive the time series feature vector, risk trend status encoding, tunneling state of the roadheader, and real-time air volume.

[0129] The time series feature vector represents the multi-scale time series features of historical air pressure changes; the risk trend status encoding can adopt 3D one-hot encoding, such as low risk [1,0,0], medium risk [0,1,0], high risk [0,0,1];

[0130] The tunneling state of the roadheader can be a binary parameter, such as 0 representing shutdown and 1 representing tunneling; the real-time air volume represents the real-time air volume value at the current moment.

[0131] All input data is aligned by timestamp to ensure synchronous processing of data within the same time window.

[0132] The feature fusion layer is used to splice the time series feature vector and the risk trend status encoding vector by channel to form a fused feature; perform sliding window mean filtering on the real-time air volume to generate an air volume trend feature; splice the fused feature, air volume trend feature, and tunneling state parameter into a joint feature vector.

[0133] Perform sliding window mean filtering on the real-time air volume sequence. For example, calculate the mean air volume within the window with a 5-minute window and a 1-minute step, and convert 30 minutes of data into a 25-dimensional trend feature to represent the recent air volume change direction, such as rising, falling, or stable.

[0134] When the tunneling state is 1, the corresponding dimension in the joint feature vector is marked as 1, triggering the model to adjust the weights of the tunneling-related features.

[0135] The gradient boosting tree ensemble layer contains a third specified number of regression trees, and each regression tree performs node splitting based on the joint feature vector and outputs the ventilation volume prediction base value for each time step.

[0136] The gradient boosting tree ensemble layer contains a specified number (such as 100) of regression trees. Each tree performs node splitting based on the joint feature vector. When splitting, calculate the information gain of each feature, and select the feature with the largest gain and the threshold to divide the data.

[0137] For example, if the 5th dimension of the air volume trend feature is greater than 0.8 and the risk encoding is high risk, the node is split into a left subtree (predicting high air volume) and a right subtree (predicting low air volume).

[0138] Each regression tree can predict the base value of the ventilation volume at a specified future time step (such as 12 time steps, 1 hour in the future, with a step of 5 minutes each). The prediction results of 100 trees form a base value matrix for subsequent interval calculations.

[0139] The output layer is used to perform quantile regression on the base value of the ventilation volume prediction. The quantile loss function is used to calculate the upper quantile and lower quantile of the specified future time range, generating a ventilation volume prediction interval.

[0140] The output layer directly predicts the upper and lower bounds of the interval using the quantile loss function. For example, setting the upper quantile to 0.95 and the lower quantile to 0.05, the model minimizes the quantile loss so that 95% of the historical true values fall within the prediction interval.

[0141] According to the ratio of the current dynamic risk index to the historical maximum risk value, the interval width is dynamically scaled. For example, when the dynamic risk index reaches 80% of the historical peak, the upper quantile is increased from 0.95 to 0.98, and the lower quantile is decreased from 0.05 to 0.02, and the interval width is expanded by 6% to cope with the uncertainty under high risk.

[0142] When the risk code is high risk, the improved GBRT architecture can automatically enhance the sensitivity to sudden changes in air pressure characteristics (such as a sudden drop in air pressure caused by blasting), and timely adjust the prediction interval.

[0143] When the tunneling state is 1, the features related to equipment operation in the regression tree (such as the tunneling state is 1 and the air volume trend is greater than 0.5) trigger a high base value prediction, which can reflect the incremental demand for ventilation volume during tunneling operations.

[0144] In some embodiments, the quantile output layer dynamically scales the upper and lower bounds of the ventilation volume prediction interval according to the ratio of the current dynamic risk index to the historical risk interval, and the scaling coefficient is positively correlated with the dynamic risk index.

[0145] Pre-statistically calculate the historical data of the dynamic risk index during the normal operation cycle of the mine and calculate its maximum value and minimum value , forming a historical risk interval , .

[0146] Map the dynamically calculated dynamic risk index to the historical interval and calculate the relative ratio Ratio, where Ratio is: .

[0147] Set the scaling coefficient to be linearly positively correlated with Ratio, and set , where Represents a preset constant, such as 0.2.

[0148] It can be obtained by analyzing the distribution of prediction errors in historical high - risk events (such as gas concentration exceeding the limit and alarming), and determining the interval expansion ratio required to cover 95% of extreme cases.

[0149] Initial upper quantile Can be set to 0.95, and the lower quantile Can be set to 0.05, and the dynamic scaling rule is: , ; where and Represent the upper quantile and lower quantile after dynamic scaling respectively.

[0150] When the upper quantile after dynamic scaling > 1 or the lower quantile after dynamic scaling < 0, it is forcibly truncated to 1.0 or 0.0 to ensure the legality of the quantile.

[0151] When the sudden increase in gas concentration causes the dynamic risk index to approach the maximum value , significantly increases, and the prediction interval expands to cover a larger range of uncertainties. For example, when the gas concentration exceeds the limit and alarms, if the dynamic risk index = 0.85, the width of the prediction interval expands from the default 0.9 to 1.0, improving safety redundancy.

[0152] When the dynamic risk index is low, such as = 0.2, decreases, and the interval width narrows to avoid energy waste caused by excessive ventilation.

[0153] In some embodiments, as Figure 2 shown, the present disclosure provides a ventilation data analysis system for intelligent mine ventilation, which includes: a multi - source environmental parameter perception module, a dynamic risk assessment and fusion module, a risk coding and barometric time - series feature generation module, and a multi - modal ventilation decision - making module.

[0154] The multi - source environmental parameter perception module is used to obtain mine tunnel environmental parameters, and the mine tunnel environmental parameters include gas concentration, real - time air volume, and temperature.

[0155] The dynamic risk assessment and fusion module is used to calculate the dynamic risk index according to the mine tunnel environmental parameters.

[0156] The risk coding and barometric pressure time series feature generation module is used to generate a risk trend status code according to the dynamic risk index; obtain the atmospheric pressure sequence within a specified past time range, and process the atmospheric pressure sequence using an adaptive atmospheric pressure model. The adaptive atmospheric pressure model is based on a temporal convolutional network, dynamically determines the optimal dilation rate of the dilated convolutional layer, and outputs a time series feature vector after integrating multi-scale information in depth;

[0157] The multi-modal ventilation decision module is used to obtain the tunneling state of the tunneling machine; according to the time series feature vector, the risk trend status code, the tunneling state of the tunneling machine, and the real-time air volume, obtain the ventilation volume prediction interval for a specified future time range.

[0158] In some embodiments, the present disclosure provides a device, which includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement the steps of a ventilation data analysis method for intelligent mine ventilation when executing the computer program.

[0159] In some embodiments, the present disclosure provides a readable storage medium, in which computer program instructions are stored. When the computer program instructions are read and run by a processor, the steps of a ventilation data analysis method for intelligent mine ventilation are executed.

[0160] Wherein, any reference to a memory, storage, database, or other medium used in the various embodiments provided by the present invention may include non-volatile and / or volatile memory. The non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. The volatile memory may include random access memory (RAM) or an external cache memory.

[0161] It should be noted that in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variation thereof is intended to cover a non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising a..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0162] Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A ventilation data analysis method for intelligent ventilation in a mine, characterized in that, Including: Obtain the mine tunnel environment parameters, which include gas concentration, real-time air volume, and temperature; Calculate the dynamic risk index based on the mine tunnel environment parameters; Generate a risk trend status code according to the dynamic risk index; Obtain the atmospheric pressure sequence within a specified past time range, and process the atmospheric pressure sequence using an adaptive atmospheric pressure model. The adaptive atmospheric pressure model is based on a temporal convolutional network, dynamically determines the optimal dilation rate of the dilated convolutional layer, and outputs a temporal feature vector after integrating multi-scale information in depth; Obtain the tunneling state of the roadheader; according to the temporal feature vector, risk trend status code, tunneling state of the roadheader, and real-time air volume, obtain the ventilation volume prediction interval for a specified future time range; Among them, obtaining the ventilation volume prediction interval for a specified future time range according to the temporal feature vector, risk trend status code, tunneling state of the roadheader, and real-time air volume includes: Use the improved GBRT architecture to process the temporal feature vector, risk trend status code, tunneling state of the roadheader, and real-time air volume, and output the ventilation volume prediction interval for a specified future time range; The improved GBRT architecture includes: an input layer, a feature fusion layer, a gradient boosting tree ensemble layer, and an output layer; The input layer is used to receive the temporal feature vector, risk trend status code, tunneling state of the roadheader, and real-time air volume; The feature fusion layer is used to splice the temporal feature vector and the risk trend status code vector by channel to form a fused feature; perform sliding window mean filtering on the real-time air volume to generate an air volume trend feature; splice the fused feature, air volume trend feature, and tunneling state parameter into a joint feature vector; The gradient boosting tree ensemble layer contains a third specified number of regression trees, and each regression tree performs node splitting based on the joint feature vector and outputs the ventilation volume prediction base value for each time step; The output layer is used to perform quantile regression on the ventilation volume prediction base value, calculate the upper quantile and lower quantile for a specified future time range using the quantile loss function, and generate the ventilation volume prediction interval.

2. The ventilation data analysis method for intelligent mine ventilation according to claim 1, wherein Obtain the mine tunnel environment parameters, including: Configure the basic acquisition parameters, including: setting the sampling frequencies of gas concentration, real-time air volume, and temperature, defining the time window and space window, as well as the sliding intervals of the time window and space window; Based on the basic acquisition parameters, generate a three-dimensional temporal tensor including a time axis, spatial node distribution, and parameter channels by sliding the time window; Dynamically crop the three-dimensional temporal tensor according to the sliding of the space window to obtain valid node data.

3. The ventilation data analysis method for intelligent mine ventilation according to claim 2, wherein Defining the space window includes: Taking the roadheader as the reference point, use an ultra-wideband positioning system to obtain the real-time coordinates of the reference point; Determine the coordinates of the sensor and calculate the actual distance between the sensor and the reference point; Filter out the sensors whose actual distance meets the preset distance.

4. The ventilation data analysis method for intelligent mine ventilation according to claim 2, wherein, Calculating the dynamic risk index according to the mine tunnel environment parameters includes: For the valid node data within the current time window and space window: calculate the mean value of the gas concentration according to the sampling frequency and spatial node distribution of the gas concentration; normalize the real-time air volume, and calculate the mean value of the normalized real-time air volume according to the sampling frequency and spatial node distribution of the real-time air volume; calculate the mean value of the temperature according to the sampling frequency and spatial node distribution of the temperature; Calculate the gas concentration risk factor, air volume risk factor, and temperature risk factor based on the mean value of the gas concentration, the mean value of the real-time air volume after normalization, and the mean value of the temperature; Calculate the dynamic risk index based on the gas concentration risk factor, air volume risk factor, and temperature risk factor.

5. The ventilation data analysis method for intelligent mine ventilation according to claim 1, characterized in that, Generate a risk trend status code according to the dynamic risk index, including: Determine two or more consecutive increasing risk levels, each risk level corresponding to a unique code; define a demarcation threshold that is one less than the number of risk levels, and all demarcation thresholds are arranged in ascending order; Determine the demarcation threshold interval where the dynamic risk index is located, and output the risk level code according to the coding mapping mechanism; Among them, the coding mapping mechanism includes: Map the dynamic risk index lower than the minimum demarcation threshold to the lowest risk level code; Map the dynamic risk index between every two adjacent demarcation thresholds to the intermediate risk level codes in the order of the demarcation threshold arrangement; Map the dynamic risk index higher than or equal to the maximum demarcation threshold to the highest risk level code.

6. The ventilation data analysis method for intelligent mine ventilation according to claim 1, characterized in that, The adaptive atmospheric pressure model includes an input layer, a first specified number of dilated convolutional layers, and an output layer; The input layer is used to receive the atmospheric pressure sequence within a specified past time range; The first specified number of dilated convolutional layers are assigned optimal dilation rates to capture the change characteristics of the atmospheric pressure at different periods, obtaining multiple feature maps, splicing the feature maps by channels, and integrating multi-scale information into a feature tensor; The output layer flattens the feature tensor to obtain a time series feature vector.

7. The ventilation data analysis method for intelligent mine ventilation according to claim 6, characterized in that In the process of assigning the optimal dilation rate to the first specified number of dilated convolutional layers, obtaining the optimal dilation rate includes: Obtain the atmospheric pressure sequence within a specified past time range; Perform multi-scale decomposition on the atmospheric pressure sequence to obtain multiple subsequences; Extract the statistical features of the subsequences to obtain feature vectors; Cluster the feature vectors of all subsequences to obtain a second specified number of cluster centers and the cluster labels to which each subsequence belongs; Classify the initial dilation rate for each cluster center to obtain a set of initial dilation rates; When training the adaptive atmospheric pressure model, calculate the importance weight of the subsequence corresponding to each cluster in the current input; Fine-tune the initial dilation rate according to the importance weight of the subsequence corresponding to each cluster in the current input to obtain the final dilation rate; Construct a set of final dilation rates, use the set of final dilation rates to train the adaptive atmospheric pressure model and feedback adjustment to obtain the optimal dilation rate.

8. The ventilation data analysis method for intelligent mine ventilation according to claim 1, wherein The quantile output layer dynamically scales the upper and lower bounds of the ventilation volume prediction interval according to the ratio of the current dynamic risk index to the historical risk interval, and the scaling coefficient is positively correlated with the dynamic risk index.

9. A ventilation data analysis system for intelligent ventilation in a mine, characterized in that, It includes: A multi-source environmental parameter perception module, which is used to obtain the mine environment parameters, and the mine environment parameters include gas concentration, real-time air volume, and temperature; A dynamic risk assessment and fusion module, which is used to calculate the dynamic risk index according to the mine environment parameters; A risk coding and atmospheric pressure time series feature generation module, which is used to generate a risk trend status code according to the dynamic risk index; Obtain the atmospheric pressure sequence within a specified past time range, and process the atmospheric pressure sequence using an adaptive atmospheric pressure model. The adaptive atmospheric pressure model is based on a temporal convolutional network, dynamically determines the optimal dilation rate of the dilated convolutional layer, and outputs a temporal feature vector after integrating multi-scale information in depth; A multi-modal ventilation decision-making module, which is used to obtain the tunneling state of the tunneling machine; according to the temporal feature vector, the risk trend state encoding, the tunneling state of the tunneling machine, and the real-time air volume, obtain the ventilation volume prediction interval for a specified future time range; Among them, obtaining the ventilation volume prediction interval for a specified future time range according to the temporal feature vector, the risk trend state encoding, the tunneling state of the tunneling machine, and the real-time air volume includes: Use the GBRT improved architecture to process the temporal feature vector, the risk trend state encoding, the tunneling state of the tunneling machine, and the real-time air volume, and output the ventilation volume prediction interval for a specified future time range; The GBRT improved architecture includes: an input layer, a feature fusion layer, a gradient boosting tree ensemble layer, and an output layer; The input layer is used to receive the temporal feature vector, the risk trend state encoding, the tunneling state of the tunneling machine, and the real-time air volume; The feature fusion layer is used to splice the temporal feature vector and the risk trend state encoding vector by channel to form a fused feature; perform sliding window mean filtering on the real-time air volume to generate an air volume trend feature; splice the fused feature, the air volume trend feature, and the tunneling state parameters into a joint feature vector; The gradient boosting tree ensemble layer contains a third specified number of regression trees, and each regression tree performs node splitting based on the joint feature vector and outputs the ventilation volume prediction base value for each time step; The output layer is used to perform quantile regression on the ventilation volume prediction base value, calculate the upper quantile and the lower quantile for a specified future time range using a quantile loss function, and generate a ventilation volume prediction interval.

Images