Coal mine hydrological dynamic monitoring system and method based on neural network learning

Abstract

The invention belongs to the technical field of coal mine hydrological monitoring, and discloses a coal mine hydrological dynamic monitoring system and method based on neural network learning. According to the method, coal mine hydrological dynamic image segmentation is carried out based on an improved neural network, monitoring indexes of a coal mine hydrological dynamic evolution curve are obtained, and classification and extraction are carried out according to collection sources of the monitoring indexes of the coal mine hydrological dynamic evolution curve; vectorization processing of qualitative evolution trend monitoring analysis, historical record aging weighting and standardization processing of original mine water level, water temperature and flow parameters are carried out; and comprehensive evaluation of dynamic monitoring accuracy: performing standardization processing, determining a principal component, determining an entropy weight, and performing judgment by using a comprehensive evolution trend monitoring analysis value calculation formula of the dynamic monitoring accuracy. According to the method, the coal mine hydrological dynamic state is read through the neural network, so that the specific state or evolution trend is judged, and the monitoring accuracy and precision are further improved.

Description

Technical Field

[0001] The invention belongs to the technical field of coal mine hydrological monitoring, and in particular relates to a coal mine hydrological dynamic monitoring system and method based on neural network learning. Background Art

[0002] During the construction and production process of coal mines, monitoring equipment needs to be installed on mining equipment for real-time dynamic monitoring. When existing equipment is used for monitoring, the vibration generated during the mining process will cause the fault water to shake, and at the same time cause cracks to appear inside the rock mass, causing the flow direction of the fault water to change. The amount of water transmitted in the cracks is small, resulting in the inability of existing monitoring equipment to dynamically monitor the flow path of the fault water in real time and accurately, and may be unable to predict the occurrence of a rock side collapse accident, posing a threat to the safe operation of the monitoring equipment.

[0003] In order to solve the above problems, the invention patent "A coal mine hydrological dynamic monitoring system" (publication number CN114412571A, publication date 20220429) discloses a coal mine hydrological dynamic monitoring system, including graphite powder for enhancing the conductivity of fault water, an injection anchor for injecting graphite powder into a large water-rich fault, and a fixed plate installed on the back side of the tunneling equipment, a shock-absorbing slide rail is installed on the side of the fixed plate away from the tunneling equipment, a monitoring mechanism is slidably connected to the surface of the shock-absorbing slide rail, and an arc-shaped slide groove is opened on the surface of the fixed plate; the tail end of the injection anchor is rotatably connected to a flap, and the injection anchor A powder storage chamber is provided inside, and the graphite powder is stored in the powder storage chamber. An inner chamber is provided on the side of the injection anchor and at the bottom of the flap. An electromagnet is fixedly installed at the bottom of the inner chamber, and the electromagnet is magnetically connected to the flap. A sealing plate is fixedly installed at the inner bottom of the inner chamber, and a transmission pipe is fixedly installed on the surface of the sealing plate. The input end of the transmission pipe is connected to the powder storage chamber, and the output end of the sealing plate is rotatably connected to a rotating nozzle. A rotating nozzle is installed on the side of the rotating nozzle away from the sealing plate. A through hole is provided for the input of the graphite powder, and the through hole is intermittently connected to the transmission pipe. The steps for use are as follows:

[0004] A: Before the tunnel section is excavated, the tunnel hydrological electromagnetic transient detection equipment installed on the back of the mining equipment detects the water-rich faults in the tunnel section to quickly find the large water-rich faults in the tunnel section, and drives the tail of the injection anchor into the large water-rich faults. The electromagnet is controlled to be intermittently powered, and the electromagnet is magnetically connected to the flap. The pull ring on the back of the flap is connected to the screw through a pull rope, which pulls the screw to move inside the sleeve. Because the screw and the gear are threaded, the rotating nozzle and the gear are driven to rotate as a whole, so that the through hole and the transmission pipe overlap intermittently. The rotating nozzle and the rotating nozzle evenly spray graphite powder into the water-rich fault to increase the conductivity of the water in the fault.

[0005] B: After completing step A, due to the mining action during the excavation process, many small faults appear on the edge of the large fault. The tunnel hydrological electromagnetic transient detection equipment detects the hydrological conditions of the small faults with increased signals, and the controller quickly identifies the extension trajectory of the small faults;

[0006] C: While completing step B, when it is detected that the water inside the small fault reaches the rock surface, the controller controls the sliding wheel inside the sliding platform to slide on the outer surface of the shock-absorbing slide rail, so that the sliding platform as a whole moves in the direction opposite to the collapse surface, providing the first layer of protection for the monitoring mechanism as a whole. At the same time, due to the movement of the sliding platform as a whole, the upper surface of the sliding platform is rotatably connected to the protective shell through the connecting piece, telescopic shaft, connecting sleeve, connecting plate and rotating rod, and the connecting plate slides in the arc-shaped slide groove, thereby driving the protective shell and the flexible pad to move in the opposite direction of the sliding platform, providing the second layer of protection for the sliding platform as a whole, so as to prevent the detection equipment from being damaged by sudden water seepage accidents;

[0007] D: While completing step C, the controller transmits detection and alarm signals to the alarm center in real time, and also sends the extension position and water volume information of the small fault to the alarm center.

[0008] Furthermore, in the traditional monitoring method, the water level (water pressure) changes are monitored by manual measurement and recording at the hydrological observation holes above and below the mine; the flow rate of pipelines and open channels is also monitored by manual measurement with instruments. The traditional monitoring method cannot monitor the required monitoring data in real time, and it relies on manual recording and management of this series of data, and it is also necessary to count and analyze the data, which will be extremely labor-intensive and prone to errors, resulting in management confusion.

[0009] Recently, with the development of machine learning, attempts have been made to automatically perform tasks such as identifying or classifying images through computer systems. In particular, attempts have been made to use a neural network, one of machine learning methods (for example, a deep learning method using a convolutional neural network (CNN)) to automate various classifications or judgments performed by people, and to read the hydrological dynamics of coal mines by using deep learning in which a neural network (for example, CNN) is used, thereby performing judgments on specific states or evolving trends, and further improving the accuracy of monitoring. Therefore, it is necessary to propose a method for monitoring the hydrological dynamics of coal mines based on neural network learning. Summary of the invention

[0010] In order to overcome the problems existing in the related art, the disclosed embodiments of the present invention provide a coal mine hydrological dynamic monitoring system and method based on neural network learning.

[0011] The technical solution is as follows: A method for dynamic monitoring of coal mine hydrology based on neural network learning, comprising:

[0012] S1, performing coal mine hydrological dynamic image segmentation based on an improved neural network to obtain monitoring indicators of a coal mine hydrological dynamic evolution curve, wherein the coal mine hydrological dynamic evolution curve includes a mine water level, water temperature and flow data curve;

[0013] S2, classify and extract the sources of monitoring index collection of coal mine hydrological dynamic evolution curve, conduct vector processing of qualitative evolution trend monitoring and analysis, weight the timeliness of historical records, and normalize the original mine water level, water temperature and flow parameters;

[0014] S3, comprehensive evaluation of dynamic monitoring accuracy, standardization processing, determination of principal components, determination of entropy weights, and judgment using the calculation formula for the comprehensive evolution trend monitoring analysis value of dynamic monitoring accuracy.

[0015] In step S1, the coal mine hydrological dynamic image segmentation is performed based on the improved neural network to obtain the monitoring indicators of the coal mine hydrological dynamic evolution curve, including:

[0016] S101, obtaining a training data set, and cutting the coal mine hydrological dynamic image and the real coal mine hydrological area marked image in the training data set, and cutting the coal mine hydrological dynamic image and the real coal mine hydrological area marked image in the training data set into a size of 512×512;

[0017] S102, constructing a coding module of an improved neural network, and using the coal mine hydrological dynamic images and the real coal mine hydrological area marked images in the cut training data set to train the constructed coding module of the improved neural network;

[0018] S103, performing light vectorization image segmentation using the trained encoding and decoding module of the improved neural network.

[0019] In step S102, the encoding and decoding module of the constructed improved neural network is trained, including:

[0020] The coding unit uses the coding chip to output 6 types of coal mine hydrological dynamic characteristic maps of 1 / 2, 1 / 4, 1 / 8, 1 / 16, 1 / 32 and 1 / 64;

[0021] The decoding unit, based on the six coal mine hydrological dynamic characteristic maps of 1 / 2, 1 / 4, 1 / 8, 1 / 16, 1 / 32 and 1 / 64 output by the encoding chip, uses the convolution structure to re-emphasize the mine water level, water temperature and flow parameters, and combines the channel attention strategy for decoding to obtain the segmentation result.

[0022] Furthermore, the decoding is combined with the channel attention strategy to obtain the segmentation results, including:

[0023] The 1 / 64 coal mine hydrological dynamic characteristic map is used as the input of the decoding unit, and the 1 / 2, 1 / 4, 1 / 8, 1 / 16, and 1 / 32 coal mine hydrological dynamic characteristic maps are used as the feature information of the jump connection and are spliced ​​and fused with the encoding unit;

[0024] The deconvolution process is used to expand the size of the coal mine hydrological dynamic feature map to four times of its original size, and the coal mine hydrological dynamic feature map is gradually restored to the original image resolution size, and then spliced ​​with the coding features output by each layer of the coding unit; the output layer is used to map the coal mine hydrological dynamic feature map into a specific number of mine water level, water temperature and flow classification types to predict the pixel mine water level, water temperature and flow classification types to obtain the segmentation result.

[0025] Further, the decoding unit comprises:

[0026] Model heavy mine water level, water temperature and flow parameterization module, multi-scale fusion convolution module and downlink acquisition module;

[0027] Modules for parameterizing water level, water temperature and flow in model heavy mines to enhance useful features;

[0028] The multi-scale fusion convolution module is used to extract features by convolution with different convolution kernel sizes, and to splice and fuse the coal mine hydrological dynamic feature maps obtained by multi-scale fusion convolution;

[0029] Downlink acquisition module, used to transmit the fused feature information to the deep network;

[0030] The multi-scale fusion convolution module is sequentially provided with three convolution kernels of 6×6, 10×10 and 14×14 respectively, three branch features for splicing processing, and a 6×6 convolution for fusion;

[0031] The downlink acquisition module comprises:

[0032] The maximum pooling downstream acquisition unit is used to change the dimension of the coal mine hydrological dynamic feature map through 2×2 convolution;

[0033] A downlink acquisition unit, used for performing downlink acquisition by using a 6×6 convolution with a step size of 4;

[0034] The fusion unit is used to fuse the maximum pooling downstream acquisition unit and the downstream acquisition unit by summing the corresponding elements, and access the SiLU function for nonlinear activation.

[0035] Step S102, using the coal mine hydrological dynamic images and the real coal mine hydrological area marked images in the cut training data set to train the encoding and decoding module of the constructed improved neural network, including:

[0036] The cosine annealing training strategy is adopted, and the coal mine hydrological dynamic image in the cut training data set is input into the encoding and decoding module of the constructed improved neural network to obtain the segmentation result map of the corresponding training image; the obtained segmentation result map is compared with the real coal mine hydrological area marked image, and the loss function loss value is calculated; it is judged whether the loss function loss value has a decreasing evolution trend, and if the loss function loss value has not evolved a decreasing trend, the model parameters are updated by back propagation; when the loss function loss value has a decreasing evolution trend, the encoding and decoding module of the trained improved neural network is obtained;

[0037] The loss function is as follows:

[0038]

[0039]

[0040]

[0041]

[0042] S all =βS SCE +ηS Dice

[0043] In the formula, S CE is the cross entropy loss, S SCE is the cross entropy loss smoothed using the coal mine hydrological region labeling, f′ ic is the real mine water level, water temperature and flow classification type of the coal mine hydrological dynamic monitoring object i after smoothing with the coal mine hydrological area markers, S Dice is the ideal loss, S all is the total loss function, N is the total number of coal mine hydrological dynamic monitoring objects, M is the number of mine water level, water temperature and flow classification types; f ic is the classification type of the real mine water level, water temperature and flow rate of the coal mine hydrological dynamic monitoring object i. If the real mine water level, water temperature and flow rate classification type of the coal mine hydrological dynamic monitoring object i is c, it is 1, otherwise it is 0; h ic is the prediction result of the coal mine hydrological dynamic monitoring object i belonging to the mine water level, water temperature and flow classification type c, δ is the hyperparameter of the smoothing amount, |A∩F| is the intersection of sets A and F, i is the coal mine hydrological dynamic monitoring object, c is the classification type, A is the number of elements in set A, F is the number of elements in set F, β is the first weighting coefficient, and η is the second weighting coefficient.

[0044] In step S2, the monitoring index collection sources of the coal mine hydrological dynamic evolution curve are classified and extracted, including:

[0045] The first category: the processing time of coal mine hydrological dynamic evolution trend, the processing time of coal mine hydrological dynamic evolution stable trend, the calculation cost of coal mine hydrological dynamic evolution stable trend, the sustainability of coal mine hydrological dynamic evolution stable trend calculation, and the availability of coal mine hydrological dynamic evolution stable trend calculation are all extracted from the monitoring image collection of each coal mine hydrological dynamic evolution area;

[0046] The second category: accuracy of calculation quality of steady trend of coal mine hydrological dynamic evolution, confirmation of calculation quality of steady trend of coal mine hydrological dynamic evolution, confirmation of normal calculation quality of coal mine hydrological dynamic evolution trend and accuracy of normal calculation quality of coal mine hydrological dynamic evolution trend, all of which are extracted from the system's use of coal mine hydrological image evolution trend monitoring and analysis.

[0047] In step S2, the historical records are weighted in terms of timeliness, including:

[0048] Taking the time parameter as the basis, the following formula is used:

[0049]

[0050] Where, k is a final value obtained after calculating all evolution trend monitoring and analysis records of the monitoring indicator K of a coal mine hydrological dynamic evolution curve, t(n) is a time weighted function, kn is the nth data of the monitoring indicator of a coal mine hydrological dynamic evolution curve, and n is the total number of data;

[0051] The expression of the time weighting function t(n) is:

[0052]

[0053] Where, T n is the time value of the i-th record in the monitoring index of the coal mine hydrological dynamic evolution curve with a total number of data being n, ΔT is the time difference, T i is the time value of the i-th record in the monitoring index of a coal mine hydrological dynamic evolution curve, in minutes; T i It is the time value of the i-th record in the monitoring index of a coal mine hydrological dynamic evolution curve, in minutes;

[0054] Standardization of original mine water level, water temperature and flow parameters, including:

[0055] First, the inverse parameters are converted into positive parameters to obtain the evolution trend monitoring analysis value set u of multiple performance dynamic monitoring accuracy coal mine hydrological dynamic evolution curve monitoring index service u = {u1, u2…un}, u i is the set of values ​​for monitoring and analyzing the evolution trend of the i-th performance;

[0056] u min=min{u1,u2…un}

[0057] u max =max{u1,u2…un}

[0058] V1=u min +u max

[0059] V2=1

[0060] In the formula, u min Minimum value for monitoring indicator service evolution trend monitoring and analysis, u max It is the maximum value of the monitoring indicator service evolution trend monitoring and analysis, V1, V2 are constants, min{} is the minimum set of evolution trend monitoring and analysis values, and max{} is the maximum set of evolution trend monitoring and analysis values;

[0061] After simplification, the following formula is obtained:

[0062] u inew =u min +u max -u i

[0063] In the formula, u inew is the value obtained after normalization of the monitoring index service u of the ith performance coal mine hydrological dynamic evolution curve, u i is the set of analysis values ​​for monitoring the evolution trend of the i-th performance.

[0064] In step S3, the dynamic monitoring accuracy is comprehensively evaluated, including:

[0065] After standardization, the expression is:

[0066]

[0067] In the formula, is the evolution trend monitoring analysis value after standardization, k je is the dynamic monitoring evolution trend analysis value of the jth performance and the eth number of coal mine hydrological dynamic evolution curves, O() is the dynamic monitoring evolution trend analysis function, is the average value of the j-th performance evolution trend monitoring analysis, d j is the jth performance evolution trend monitoring and analysis coefficient, and e is the eth item of data;

[0068] The variable parameter relationship matrix D is established as:

[0069] D=(g ij ) m×m×m

[0070] In the formula, m×m×m is the combination of mine water level, water temperature and flow parameters;

[0071]

[0072] In the formula, g ij is the variable value of the combination of the i-th performance and the j-th performance, l ij is the evolution trend value of the combination of the i-th performance and the j-th performance, l ii is the evolution trend value of multiple i performance overlaps, l jj is the evolution trend value of multiple j performance overlaps, k i is the i-th performance evolution trend monitoring and analysis value, is the average value of the i-th performance evolution trend monitoring analysis, k j is the jth performance evolution trend monitoring and analysis value, The average value of the j-th performance evolution trend monitoring and analysis;

[0073] Find the characteristic root and corresponding eigenvector of D, let the characteristic root a1≥a2…am≥0, and the corresponding eigenvector be c1≥c2…cm;

[0074] Determine the principal components: the jth principal component W j The cumulative variance contribution rate of is:

[0075]

[0076] In the formula, Γ n is the variance contribution rate, n is the total number of j performances, m is one of the mine water level, water temperature and flow parameters, a j is the j performance variance contribution rate;

[0077] Γ n When >70%, take the first m principal components W;

[0078] Determine the entropy weight, calculate the comprehensive evolution trend monitoring and analysis value of dynamic monitoring accuracy, obtain the main component W = {W1, W2...Wm} and the corresponding characteristic root a = {a1≥a2...am}, obtain the comprehensive evolution trend monitoring and analysis value of dynamic monitoring accuracy, and judge the system performance.

[0079] Another object of the present invention is to provide a coal mine hydrological dynamic monitoring system based on neural network learning, the system comprising:

[0080] A monitoring index acquisition module for the coal mine hydrological dynamic evolution curve is used to perform coal mine hydrological dynamic image segmentation based on an improved neural network to acquire monitoring indicators of the coal mine hydrological dynamic evolution curve, wherein the coal mine hydrological dynamic evolution curve includes a mine water level, water temperature and flow data curve;

[0081] The processing module is used to classify and extract the sources of monitoring indicators of the coal mine hydrological dynamic evolution curve, perform vector processing of qualitative evolution trend monitoring and analysis, weight the timeliness of historical records, and normalize the original mine water level, water temperature and flow parameters;

[0082] The monitoring and analysis module is used for comprehensive evaluation of dynamic monitoring accuracy, standardization processing, determination of principal components, determination of entropy weights, and judgment using the calculation formula for the comprehensive evolution trend monitoring and analysis value of dynamic monitoring accuracy.

[0083] In combination with all the above-mentioned technical solutions, the beneficial effects of the present invention are as follows: the present invention automates various classifications or judgments performed by people through neural networks (for example, using deep learning methods of neural networks), and reads the dynamics of coal mine hydrology (mine water level, water temperature and flow data) by using deep learning using neural networks (for example, CNN), thereby executing judgments on specific states or evolutionary trends, thereby further improving the accuracy and precision of monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] The accompanying drawings herein are incorporated in and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description, serve to explain the principles of the present disclosure;

[0085] Figure 1 is a diagram of a method for dynamic monitoring of coal mine hydrology based on neural network learning provided by an embodiment of the present invention;

[0086] Figure 2 The embodiment of the present invention provides a flow chart of monitoring indicators for segmenting coal mine hydrological dynamic images based on an improved neural network to obtain a coal mine hydrological dynamic evolution curve;

[0087] Figure 3 Schematic diagram of a coal mine hydrological dynamic monitoring system based on neural network learning provided by an embodiment of the present invention;

[0088] Figure 4 This is a real coal mine hydrological area marking image and a model segmentation result rendering of the present invention;

[0089] In the figure: 1. Monitoring index acquisition module of coal mine hydrological dynamic evolution curve; 2. Processing module; 3. Monitoring and analysis module. DETAILED DESCRIPTION

[0090] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention, so the present invention is not limited by the specific implementation disclosed below.

[0091] Embodiment 1, as Figure 1 As shown, the method for dynamic monitoring of coal mine hydrology based on neural network learning provided by the embodiment of the present invention includes:

[0092] S1, performing coal mine hydrological dynamic image segmentation based on an improved neural network to obtain monitoring indicators of a coal mine hydrological dynamic evolution curve, wherein the coal mine hydrological dynamic evolution curve includes a mine water level, water temperature and flow data curve;

[0093] S2, classify and extract the sources of monitoring index collection of coal mine hydrological dynamic evolution curve, conduct vector processing of qualitative evolution trend monitoring and analysis, weight the timeliness of historical records, and normalize the original mine water level, water temperature and flow parameters;

[0094] S3, comprehensive evaluation of dynamic monitoring accuracy, standardization processing, determination of principal components, determination of entropy weights, and judgment using the calculation formula for the comprehensive evolution trend monitoring analysis value of dynamic monitoring accuracy.

[0095] Example 2, as another implementation of the present invention, steps S2-S3 of the coal mine hydrological dynamic monitoring method based on neural network learning of the present invention include:

[0096] 1. Dynamic monitoring accuracy algorithm Construction of monitoring index system for coal mine hydrological dynamic evolution curve;

[0097] The algorithm designed by the present invention mainly conducts multi-angle analysis from several aspects, including event processing time, quality, cost of calculating the steady trend of the dynamic evolution of coal mine hydrology, sustainability of calculating the steady trend of the dynamic evolution of coal mine hydrology, and availability of calculating the steady trend of the dynamic evolution of coal mine hydrology. Event processing time can be divided into coal mine hydrological dynamic evolution trend processing time and coal mine hydrological dynamic evolution steady trend processing time, and quality can be divided into coal mine hydrological dynamic evolution steady trend calculation quality confirmation, coal mine hydrological dynamic evolution steady trend calculation quality accuracy, coal mine hydrological dynamic evolution trend normal calculation quality confirmation, coal mine hydrological dynamic evolution trend normal calculation quality accuracy.

[0098] Based on the above description, after adding the information security mechanism, the monitoring index set of the coal mine hydrological dynamic evolution curve of the system's service quality evaluation algorithm is analyzed, and the present invention constructs the following Figure 1The three-layer decision model of the dynamic monitoring accuracy evaluation algorithm shown in the figure is Figure 1 The middle scheme layer is different security schemes formed by adding different information security mechanisms; the criterion layer is the monitoring indicator layer of the dynamic monitoring accuracy of coal mine hydrological dynamic evolution curve. According to the evolution trend monitoring and analysis values ​​of the indicators in the criterion layer, the dynamic monitoring accuracy of different security schemes is comprehensively monitored and analyzed to obtain the comprehensive evolution trend monitoring and analysis values; the target layer compares the comprehensive evolution trend monitoring and analysis values ​​of the dynamic monitoring accuracy of different security schemes to obtain the optimal security scheme.

[0099] 2. Comprehensive evaluation and solution steps of dynamic monitoring accuracy algorithm;

[0100] The dynamic monitoring accuracy algorithm is comprehensively evaluated based on the historical records of the monitoring indicators of the above-mentioned coal mine hydrological dynamic evolution curves. In order to obtain ideal evaluation results, the present invention needs to pre-process the collected historical records, and then bring them into the algorithm for comprehensive evaluation and solution to select the optimal solution. The main contents of the evaluation algorithm of the present invention are divided into two parts. The first part is the processing of parameters, and the second part is the solution process. The specific contents are as follows:

[0101] 1) Data collection and processing;

[0102] According to the classification of the collection sources of the monitoring indicators of the above-mentioned coal mine hydrological dynamic evolution curves, they can be divided into two categories: the first category is: coal mine hydrological dynamic evolution trend processing time, coal mine hydrological dynamic evolution stable trend processing time, coal mine hydrological dynamic evolution stable trend calculation cost, coal mine hydrological dynamic evolution stable trend calculation sustainability, coal mine hydrological dynamic evolution stable trend calculation availability are extracted from the monitoring image collection of each coal mine hydrological dynamic evolution area; the second category is: coal mine hydrological dynamic evolution stable trend calculation quality accuracy, coal mine hydrological dynamic evolution stable trend calculation quality confirmation, coal mine hydrological dynamic evolution trend normal calculation quality confirmation and coal mine hydrological dynamic evolution trend normal calculation quality accuracy are extracted from the system using coal mine hydrological image evolution trend monitoring and analysis.

[0103] The monitoring index of the coal mine hydrological dynamic evolution curve of the dynamic monitoring accuracy algorithm of the present invention is divided into two types according to the evolution trend monitoring and analysis method: quantitative evolution trend monitoring and analysis and qualitative evolution trend monitoring and analysis. In addition, since the historical record data is affected by the time factor, the timeliness of the record needs to be considered in data processing to ensure the accuracy of the evaluation results.

[0104] ① Vector processing of qualitative evolution trend monitoring and analysis;

[0105] Among the monitoring indicators of the second type of coal mine hydrological dynamic evolution curve, such as "coal mine hydrological dynamic evolution stable trend calculation quality accuracy", etc., the system usually uses coal mine hydrological images to perform evolution trend monitoring and analysis on this service after the use of this system service. The present invention divides these coal mine hydrological image evolution trend monitoring and analysis into five options: "very satisfied", "relatively satisfied", "generally satisfied", "unsatisfied" and "very unsatisfied". The level and vectorization interval value corresponding to each option are shown in Table 1 below.

[0106] Table 1 Qualitative evolution trend monitoring and analysis level vectorization relationship

[0107] Evolution trend monitoring analysis option Evolution trend monitoring and analysis level Vectorized value interval Very satisfied I (8,9] Quite satisfied II (6,7] Generally satisfied III (4,5] Dissatisfied IV (2,3] Very dissatisfied V [0,1]

[0108] In step S2, the historical record time weighting includes:

[0109] Since the time of the historical evolution trend monitoring and analysis records is different, two adjacent evolution trend monitoring and analysis with different time intervals will affect the final evolution trend monitoring and analysis results to a certain extent. In order to reduce the impact of different evolution trend monitoring and analysis with time difference on the results, the present invention takes the time parameter as the benchmark and designs the following formula:

[0110]

[0111] Where, k is a final value obtained after calculating all evolution trend monitoring and analysis records of the monitoring indicator K of a coal mine hydrological dynamic evolution curve, t(n) is a time weighted function, kn is the nth data of the monitoring indicator of a coal mine hydrological dynamic evolution curve, and n is the total number of data;

[0112] The expression of the time weighting function t(n) is:

[0113]

[0114] Where, T n is the time value of the i-th record in the monitoring index of the coal mine hydrological dynamic evolution curve with a total number of data being n, ΔT is the time difference, T i is the time value of the i-th record in the monitoring index of a coal mine hydrological dynamic evolution curve, in minutes; T i It is the time value of the i-th record in the monitoring index of a coal mine hydrological dynamic evolution curve, in minutes;

[0115] Combining the above two formulas, the final values ​​of the monitoring indicators of various coal mine hydrological dynamic evolution curves can be calculated, which are also the original mine water level, water temperature and flow parameters of this evaluation algorithm. Next, the original mine water level, water temperature and flow parameters are standardized.

[0116] Standardization of original mine water level, water temperature and flow parameters;

[0117] In the present invention, the original mine water level, water temperature and flow parameters are divided into two categories: positive parameters and inverse parameters, where the positive parameter means that the larger the parameter value, the better the evaluation result of the system performance, and the inverse parameter is the opposite. In the present invention, indicators such as "sustainability of coal mine hydrological dynamic evolution steady trend calculation", "availability of coal mine hydrological dynamic evolution steady trend calculation" belong to positive parameters, while indicators such as "coal mine hydrological dynamic evolution trend processing time" and "coal mine hydrological dynamic evolution steady trend calculation cost" belong to inverse parameters. The present invention needs to convert the inverse parameter into a positive parameter first. Here, the existing differential transformation formula is used: u new =V1-V2u old , where V1, V2 are constants, u old is the original mine water level, water temperature and flow parameters, u new The initial parameters are obtained after normalization.

[0118] After the above steps, the inverse parameters are first converted into positive parameters to obtain the evolution trend monitoring analysis value set u of the monitoring index service u of the dynamic evolution curve of the coal mine hydrology with multiple performance dynamic monitoring accuracy u={u1,u2…un},u i is the set of values ​​for monitoring and analyzing the evolution trend of the i-th performance;

[0119] u min =min{u1,u2…un}

[0120] u max =max{u1,u2…un}

[0121] V1=u min +u max

[0122] V2=1

[0123] In the formula, u min Minimum value for monitoring indicator service evolution trend monitoring and analysis, u max It is the maximum value of the monitoring indicator service evolution trend monitoring and analysis, V1, V2 are constants, min{} is the minimum set of evolution trend monitoring and analysis values, and max{} is the maximum set of evolution trend monitoring and analysis values;

[0124] After simplification, the following formula is obtained:

[0125] u inew =u min +u max -u i

[0126] In the formula, u inewis the value obtained after normalization of the monitoring index service u of the ith performance coal mine hydrological dynamic evolution curve, u i is the set of analysis values ​​for monitoring the evolution trend of the i-th performance.

[0127] 2) Design of comprehensive evaluation algorithm for dynamic monitoring accuracy;

[0128] In the criterion layer of the present invention, a large number of indicators are selected. When evaluating the system performance, the present invention uses the principal component analysis method to screen the indicators and analyze the indicators that have a greater impact on the evaluation results.

[0129] Data standardization. Assume that the performance set involved in the evaluation is l e ∈(l1,l2…l m ), the original mine water level, water temperature and flow parameter set m j ∈(m1,m2…m n ), the initial performance parameter matrix obtained after vectorization and normalization is shown in Table 2:

[0130] Table 2 Performance initial parameter matrix

[0131]

[0132] Due to the different nature of the evaluation indicators, the values ​​taken will also be different, so Table 2 should be standardized, and the processing formula is as follows:

[0133] Standardization is performed as follows:

[0134]

[0135] In the formula, is the evolution trend monitoring analysis value after standardization, k je is the dynamic monitoring evolution trend analysis value of the jth performance and the eth number of coal mine hydrological dynamic evolution curves, O() is the dynamic monitoring evolution trend analysis function, is the average value of the j-th performance evolution trend monitoring analysis, d j is the jth performance evolution trend monitoring and analysis coefficient, and e is the eth item of data;

[0136] The variable parameter relationship matrix D is established as:

[0137] D=(g ij ) m×m×m

[0138] In the formula, m×m×m is the combination of mine water level, water temperature and flow parameters;

[0139]

[0140] In the formula, g ij is the variable value of the combination of the i-th performance and the j-th performance, l ij is the evolution trend value of the combination of the i-th performance and the j-th performance, l ii is the evolution trend value of multiple i performance overlaps, l jj is the evolution trend value of multiple j performance overlaps, k i is the i-th performance evolution trend monitoring and analysis value, is the average value of the i-th performance evolution trend monitoring analysis, k j is the jth performance evolution trend monitoring and analysis value, The average value of the j-th performance evolution trend monitoring and analysis;

[0141] Find the characteristic root and corresponding eigenvector of D, let the characteristic root a1≥a2…am≥0, and the corresponding eigenvector be c1≥c2…cm;

[0142] Determine the principal components: the jth principal component W j The cumulative variance contribution rate of is:

[0143]

[0144] In the formula, Γ n is the variance contribution rate, n is the total number of j performances, m is one of the mine water level, water temperature and flow parameters, a j is the j performance variance contribution rate;

[0145] Γ n When >70%, take the first m principal components W;

[0146] Determine the entropy weight, calculate the comprehensive evolution trend monitoring and analysis value of dynamic monitoring accuracy, obtain the main component W = {W1, W2...Wm} and the corresponding characteristic root a = {a1≥a2...am}, obtain the comprehensive evolution trend monitoring and analysis value of dynamic monitoring accuracy, and judge the system performance.

[0147] Based on the problems existing in the prior art, the present invention integrates multiple key factors affecting system performance and designs a dynamic monitoring accuracy algorithm for scientifically and effectively evaluating system performance. The algorithm constructs a collaborative system based on the different information security evolutions added in different fields or scenarios.

[0148] It can be applied to multiple scenarios and multiple elements. For different scenarios and elements, users can preset various evaluation parameters of the system according to their needs, conduct comprehensive evolution trend monitoring and analysis of system performance from multi-dimensional modeling, and ensure that after adding information security evolution, the impact on the performance of the original system is minimized.

[0149] Example 3, as another embodiment of the present invention, Figure 2 As shown, step S1 includes:

[0150] S101, obtaining a training data set, and cutting the coal mine hydrological dynamic image and the real coal mine hydrological area marked image in the training data set, and cutting the coal mine hydrological dynamic image and the real coal mine hydrological area marked image in the training data set into a size of 512×512;

[0151] S102, constructing a coding module of an improved neural network, and using the coal mine hydrological dynamic images and the real coal mine hydrological area marked images in the cut training data set to train the constructed coding module of the improved neural network;

[0152] S103, performing light vectorization image segmentation using the trained encoding and decoding module of the improved neural network.

[0153] The embodiment of the present invention provides a method for cutting the coal mine hydrological dynamic image in the training data set and the real coal mine hydrological area marked image, including:

[0154] The coal mine hydrological dynamic images in the training data set and the real coal mine hydrological area labeled images are cut into a size of 512×512.

[0155] The improved neural network encoding and decoding module provided by the embodiment of the present invention includes:

[0156] The coding unit is used to output six types of coal mine hydrological dynamic characteristic diagrams of 1 / 2, 1 / 4, 1 / 8, 1 / 16, 1 / 32 and 1 / 64 by using a coding chip;

[0157] The decoding unit is used to decode the six coal mine hydrological dynamic characteristic maps of 1 / 2, 1 / 4, 1 / 8, 1 / 16, 1 / 32 and 1 / 64 based on the output of the encoding chip using the convolution structure heavy mine water level, water temperature and flow parameter method and combined with the channel attention strategy to obtain the segmentation result.

[0158] The embodiment of the present invention provides six kinds of coal mine hydrological dynamic characteristic maps based on the output of the coding chip 1 / 2, 1 / 4, 1 / 8, 1 / 16 and 1 / 32, 1 / 64, using the convolution structure to re-encode the mine water level, water temperature and flow parameter method and combining the channel attention strategy for decoding, including:

[0159] The 1 / 64 coal mine hydrological dynamic characteristic map is used as the input of the decoding unit, and the 1 / 2, 1 / 4, 1 / 8, 1 / 16, and 1 / 32 coal mine hydrological dynamic characteristic maps are used as the feature information of the jump connection and are spliced ​​and fused with the decoding unit;

[0160] Deconvolution processing is used to expand the size of the coal mine hydrological dynamic feature map to four times the original size and gradually restore the coal mine hydrological dynamic feature map to the original image resolution size, and then spliced ​​with the coding features output by each layer of the coding unit; and the output layer is used to map the coal mine hydrological dynamic feature map into a specific number of mine water level, water temperature and flow classification types to predict the pixel mine water level, water temperature and flow classification types to obtain the segmentation result.

[0161] The decoding unit provided in the embodiment of the present invention includes:

[0162] Model heavy mine water level, water temperature and flow parameterization module, multi-scale fusion convolution module and downlink acquisition module;

[0163] Modules for parameterizing water level, water temperature and flow in model heavy mines to enhance useful features;

[0164] The multi-scale fusion convolution module is used to extract features by convolution with different convolution kernel sizes, and to splice and fuse the coal mine hydrological dynamic feature maps obtained by multi-scale fusion convolution;

[0165] The downlink acquisition module is used to transmit the fused feature information to the deep network.

[0166] The multi-scale fusion convolution module provided by the embodiment of the present invention is sequentially provided with three convolution kernels with convolution kernel sizes of 6×6, 10×10 and 14×14 respectively, three branch features for splicing processing, and a 6×36 convolution for fusion.

[0167] The downlink acquisition module provided in the embodiment of the present invention includes:

[0168] The maximum pooling downstream acquisition unit is used to change the dimension of the coal mine hydrological dynamic feature map through 2×2 convolution;

[0169] A downlink acquisition unit, used for performing downlink acquisition by using a 6×6 convolution with a step size of 4;

[0170] The fusion unit is used to fuse the maximum pooling downstream acquisition unit and the downstream acquisition unit by summing the corresponding elements, and access the SiLU function for nonlinear activation.

[0171] The embodiment of the present invention provides a method for training the encoding and decoding module of the constructed improved neural network using the coal mine hydrological dynamic image and the real coal mine hydrological area marked image in the cut training data set, including:

[0172] The cosine annealing training strategy is adopted to input the coal mine hydrological dynamic image in the cut training data set into the encoding and decoding module of the constructed improved neural network to obtain the segmentation result map of the corresponding training image; the obtained segmentation result map is compared with the real coal mine hydrological area marked image, and the loss function loss value is calculated; it is judged whether the evolution trend of the loss function loss value is decreasing, if the loss function loss value does not evolve in a decreasing trend, the model parameters are updated through back propagation; when the loss function loss value evolves in a decreasing trend, the encoding and decoding module of the trained improved neural network is obtained.

[0173] The loss function is as follows:

[0174]

[0175]

[0176]

[0177]

[0178] S all =βS SCE +ηS Dice

[0179] In the formula, S CE is the cross entropy loss, S SCE is the cross entropy loss smoothed using the coal mine hydrological region labeling, f′ ic is the real mine water level, water temperature and flow classification type of the coal mine hydrological dynamic monitoring object i after smoothing with the coal mine hydrological area markers, S Dice is the ideal loss, S all is the total loss function, N is the total number of coal mine hydrological dynamic monitoring objects, M is the number of mine water level, water temperature and flow classification types; fi c is the classification type of the real mine water level, water temperature and flow rate of the coal mine hydrological dynamic monitoring object i. If the real mine water level, water temperature and flow rate classification type of the coal mine hydrological dynamic monitoring object i is c, it is 1, otherwise it is 0; h ic is the prediction result of the coal mine hydrological dynamic monitoring object i belonging to the mine water level, water temperature and flow classification type c, δ is the hyperparameter of the smoothing amount, |A∩F| is the intersection of sets A and F, i is the coal mine hydrological dynamic monitoring object, c is the classification type, A is the number of elements in set A, F is the number of elements in set F, β is the first weighting coefficient, and η is the second weighting coefficient.

[0180] The present invention is based on deep learning, uses images as input, enhances feature extraction capabilities, and improves training and reasoning efficiency; in the decoding part, a convolutional structure is used to re-calculate the mine water level, water temperature, and flow parameter method and combined with a channel attention strategy, a multi-branch structure is used during training, and the multi-branch structure is fused during reasoning, thereby improving network performance without increasing reasoning time; a multi-scale fusion convolution module is combined to obtain a coal mine hydrological dynamic feature map containing more feature information of different scales, and the feature information is transmitted to the deep network through a downlink acquisition module, thereby improving the network's ability to extract target features of different scales and better combining context information. Experimental results show that the method of the present invention improves image segmentation performance while also improving segmentation efficiency.

[0181] The present invention takes into account segmentation efficiency while improving image segmentation performance. The present invention is based on deep learning, uses images as input, uses coding units, improves feature extraction capabilities and is more light on vectorization; uses convolutional structures to train multi-branch structures for heavy mine water level, water temperature and flow parameter methods, enhances network training performance, fuses multi-branch structures into single-path structures during reasoning, improves reasoning efficiency, and uses channel attention strategies in branches to strengthen useful features and suppress invalid information; combines multi-scale fusion convolution modules in the network coding part to fuse multi-scale features, uses downlink acquisition modules to pass shallow information to deep networks, enhances the network's ability to extract features of targets of different scales, and better combines context information.

[0182] Table 3 Comparison of various indicators of model results

[0183] Method Accuracy Original Neural Network 85.99％ The present invention 88.56％

[0184] Table 4 Model light vectorization comparison

[0185] Method Training time Test time / s Original Neural Network 18h 10.71 The present invention 5h 4.52

[0186] Through the above implementation process, the model is divided into the real coal mine hydrological area marked image and the model segmentation result from top to bottom. Figure 4 , combined with Figure 4 The experimental results in Table 3 and Table 4 show that the method of the present invention has higher segmentation performance and is more light on vectorization, which shows that the method proposed by the present invention has significant effect.

[0187] Embodiment 4, as Figure 3 As shown, the coal mine hydrological dynamic monitoring system based on neural network learning provided by the embodiment of the present invention includes:

[0188] A monitoring index acquisition module 1 for the coal mine hydrological dynamic evolution curve is used to perform coal mine hydrological dynamic image segmentation based on an improved neural network to acquire monitoring indicators for the coal mine hydrological dynamic evolution curve, wherein the coal mine hydrological dynamic evolution curve includes a mine water level, water temperature and flow data curve;

[0189] Processing module 2 is used to classify and extract the sources of monitoring indicators of the coal mine hydrological dynamic evolution curve, perform vector processing of qualitative evolution trend monitoring and analysis, weight the timeliness of historical records, and normalize the original mine water level, water temperature and flow parameters;

[0190] The monitoring and analysis module 3 is used for comprehensive evaluation of dynamic monitoring accuracy, performs standardization processing, determines the principal components, determines the entropy weights, and makes judgments using the calculation formula for the comprehensive evolution trend monitoring and analysis value of dynamic monitoring accuracy.

[0191] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0192] Since the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of the present invention, their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0193] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated in one processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment.

[0194] An embodiment of the present invention further provides a computer device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor implements the steps in any of the above-mentioned method embodiments when executing the computer program.

[0195] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned method embodiments can be implemented.

[0196] An embodiment of the present invention also provides an information data processing terminal, which, when implemented on an electronic device, provides a coal mine hydrological image input interface to implement the steps in the above-mentioned method embodiments. The information data processing terminal is not limited to mobile phones, computers, and switches.

[0197] The embodiment of the present invention further provides a server, which is used to provide a coal mine hydrological image input interface to implement the steps in the above-mentioned method embodiments when executed on an electronic device.

[0198] An embodiment of the present invention provides a computer program product. When the computer program product runs on an electronic device, the electronic device can implement the steps in the above-mentioned method embodiments when executing the computer program product.

[0199] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device that can carry the computer program code to the camera / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. For example, a USB flash drive, a mobile hard disk, a disk or an optical disk.

[0200] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for dynamic monitoring of coal mine hydrology based on neural network learning, characterized in that: The method includes: S1, performing coal mine hydrological dynamic image segmentation based on an improved neural network to obtain monitoring indicators of a coal mine hydrological dynamic evolution curve, wherein the coal mine hydrological dynamic evolution curve includes a mine water level, water temperature and flow data curve; S2, classify and extract the sources of monitoring index collection of coal mine hydrological dynamic evolution curve, conduct vector processing of qualitative evolution trend monitoring and analysis, weight the timeliness of historical records, and normalize the original mine water level, water temperature and flow parameters; S3, comprehensive evaluation of dynamic monitoring accuracy, standardization processing, determination of principal components, determination of entropy weights, and judgment using the calculation formula for the comprehensive evolution trend monitoring analysis value of dynamic monitoring accuracy.

2. The method for dynamic monitoring of coal mine hydrology based on neural network learning according to claim 1 is characterized in that: In step S1, the coal mine hydrological dynamic image segmentation is performed based on the improved neural network to obtain the monitoring indicators of the coal mine hydrological dynamic evolution curve, including: S101, obtaining a training data set, and cutting the coal mine hydrological dynamic image and the real coal mine hydrological area marked image in the training data set, and cutting the coal mine hydrological dynamic image and the real coal mine hydrological area marked image in the training data set into a size of 512×512; S102, constructing a coding module of an improved neural network, and using the coal mine hydrological dynamic images and the real coal mine hydrological area marked images in the cut training data set to train the constructed coding module of the improved neural network; S103, performing light vectorization image segmentation using the trained encoding and decoding module of the improved neural network.

3. The method for dynamic monitoring of coal mine hydrology based on neural network learning according to claim 1 is characterized in that: In step S102, the encoding and decoding module of the constructed improved neural network is trained, including: The coding unit uses the coding chip to output 6 types of coal mine hydrological dynamic characteristic maps of 1 / 2, 1 / 4, 1 / 8, 1 / 16, 1 / 32 and 1 / 64; The decoding unit, based on the six coal mine hydrological dynamic characteristic maps of 1 / 2, 1 / 4, 1 / 8, 1 / 16, 1 / 32 and 1 / 64 output by the encoding chip, uses the convolution structure to re-emphasize the mine water level, water temperature and flow parameters, and combines the channel attention strategy for decoding to obtain the segmentation result.

4. The method for dynamic monitoring of coal mine hydrology based on neural network learning according to claim 3 is characterized in that: Combined with the channel attention strategy for decoding, the segmentation results are obtained, including: The 1 / 64 coal mine hydrological dynamic characteristic map is used as the input of the decoding unit, and the 1 / 2, 1 / 4, 1 / 8, 1 / 16, and 1 / 32 coal mine hydrological dynamic characteristic maps are used as the feature information of the jump connection and are spliced ​​and fused with the encoding unit; Deconvolution processing is used to expand the size of the coal mine hydrological dynamic feature map to four times the original size, and the coal mine hydrological dynamic feature map is gradually restored to the original image resolution size, and spliced ​​with the coding features output by each layer of the coding unit; the output layer is used to map the coal mine hydrological dynamic feature map into a specific number of mine water level, water temperature and flow classification types to predict the pixel mine water level, water temperature and flow classification types to obtain the segmentation result.

5. The method for dynamic monitoring of coal mine hydrology based on neural network learning according to claim 4 is characterized in that: A decoding unit, comprising: Model heavy mine water level, water temperature and flow parameterization module, multi-scale fusion convolution module and downlink acquisition module; Modules for parameterizing water level, water temperature and flow in model heavy mines to enhance useful features; The multi-scale fusion convolution module is used to extract features by convolution with different convolution kernel sizes, and to splice and fuse the coal mine hydrological dynamic feature maps obtained by multi-scale fusion convolution; Downstream acquisition module, used to transmit the fused feature information to the deep network; The multi-scale fusion convolution module is sequentially provided with three convolution kernels of 6×6, 10×10 and 14×14 respectively, three branch features for splicing processing, and a 6×6 convolution for fusion; The downlink acquisition module comprises: The maximum pooling downstream acquisition unit is used to change the dimension of the coal mine hydrological dynamic feature map through 2×2 convolution; A downlink acquisition unit, used for performing downlink acquisition by using a 6×6 convolution with a step size of 4; The fusion unit is used to fuse the maximum pooling downstream acquisition unit and the downstream acquisition unit by summing the corresponding elements, and access the SiLU function for nonlinear activation.

6. The method for dynamic monitoring of coal mine hydrology based on neural network learning according to claim 2 is characterized in that: Step S102, using the coal mine hydrological dynamic images and the real coal mine hydrological area marked images in the cut training data set to train the encoding and decoding module of the constructed improved neural network, including: The cosine annealing training strategy is adopted, and the coal mine hydrological dynamic image in the cut training data set is input into the encoding and decoding module of the constructed improved neural network to obtain the segmentation result map of the corresponding training image; the obtained segmentation result map is compared with the real coal mine hydrological area marked image, and the loss function loss value is calculated; it is judged whether the loss function loss value has a decreasing evolution trend, and if the loss function loss value has not evolved a decreasing trend, the model parameters are updated by back propagation; when the loss function loss value has a decreasing evolution trend, the encoding and decoding module of the trained improved neural network is obtained; The loss function is as follows: S all =βS SCE +ηS Dice In the formula, S CE is the cross entropy loss, S SCE is the cross entropy loss smoothed using the coal mine hydrological region labeling, f' ic is the real mine water level, water temperature and flow classification type of the coal mine hydrological dynamic monitoring object i after smoothing with the coal mine hydrological area markers, S Dice is the ideal loss, S all is the total loss function, N is the total number of coal mine hydrological dynamic monitoring objects, M is the number of mine water level, water temperature and flow classification types; f ic is the classification type of the real mine water level, water temperature and flow rate of the coal mine hydrological dynamic monitoring object i. If the real mine water level, water temperature and flow rate classification type of the coal mine hydrological dynamic monitoring object i is c, it is 1, otherwise it is 0; h ic is the prediction result of the coal mine hydrological dynamic monitoring object i belonging to the mine water level, water temperature and flow classification type c, δ is the hyperparameter of the smoothing amount, |A∩F| is the intersection of sets A and F, i is the coal mine hydrological dynamic monitoring object, c is the classification type, A is the number of elements in set A, F is the number of elements in set F, β is the first weighting coefficient, and η is the second weighting coefficient.

7. The method for dynamic monitoring of coal mine hydrology based on neural network learning according to claim 1, characterized in that: In step S2, the monitoring index collection sources of the coal mine hydrological dynamic evolution curve are classified and extracted, including: The first category: the processing time of coal mine hydrological dynamic evolution trend, the processing time of coal mine hydrological dynamic evolution stable trend, the calculation cost of coal mine hydrological dynamic evolution stable trend, the sustainability of coal mine hydrological dynamic evolution stable trend calculation, and the availability of coal mine hydrological dynamic evolution stable trend calculation are all extracted from the monitoring image collection of each coal mine hydrological dynamic evolution area; The second category: accuracy of calculation quality of steady trend of coal mine hydrological dynamic evolution, confirmation of calculation quality of steady trend of coal mine hydrological dynamic evolution, confirmation of normal calculation quality of coal mine hydrological dynamic evolution trend and accuracy of normal calculation quality of coal mine hydrological dynamic evolution trend, all of which are extracted from the system's use of coal mine hydrological image evolution trend monitoring and analysis.

8. The method for dynamic monitoring of coal mine hydrology based on neural network learning according to claim 1, characterized in that: In step S2, the historical records are weighted in terms of timeliness, including: Taking the time parameter as the basis, the following formula is used: Where, k is a final value obtained after calculating all evolution trend monitoring and analysis records of the monitoring indicator K of a coal mine hydrological dynamic evolution curve, t(n) is a time weighted function, kn is the nth data of the monitoring indicator of a coal mine hydrological dynamic evolution curve, and n is the total number of data; The expression of the time weighting function t(n) is: Where, T n is the time value of the i-th record in the monitoring index of the coal mine hydrological dynamic evolution curve with a total number of data being n, ΔT is the time difference, T i It is the time value of the i-th record in the monitoring index of a coal mine hydrological dynamic evolution curve, in minutes; Standardization of original mine water level, water temperature and flow parameters, including: First, the inverse parameters are converted into positive parameters to obtain the evolution trend monitoring analysis value set u of multiple performance dynamic monitoring accuracy coal mine hydrological dynamic evolution curve monitoring index service u = {u1,u2…un}, u i is the set of values ​​for monitoring and analyzing the evolution trend of the i-th performance; u min =min{u1,u2…un} the max =max{u1,u2…un} V1=u min +u max V2＝1 In the formula, u min Minimum value for monitoring indicator service evolution trend monitoring and analysis, u max It is the maximum value of the monitoring indicator service evolution trend monitoring and analysis, V1, V2 are constants, min{} is the minimum set of evolution trend monitoring and analysis values, and max{} is the maximum set of evolution trend monitoring and analysis values; After simplification, the following formula is obtained: in inew =in min +in max -in i In the formula, u inew is the value obtained after normalization of the monitoring index service u of the ith performance coal mine hydrological dynamic evolution curve, u i is the set of analysis values ​​for monitoring the evolution trend of the i-th performance.

9. The method for dynamic monitoring of coal mine hydrology based on neural network learning according to claim 1, characterized in that: In step S3, the dynamic monitoring accuracy is comprehensively evaluated, including: After standardization, the expression is: In the formula, is the evolution trend monitoring analysis value after standardization, k je is the dynamic monitoring evolution trend analysis value of the jth performance and the eth number of coal mine hydrological dynamic evolution curves, O( ) is the dynamic monitoring evolution trend analysis function, is the average value of the j-th performance evolution trend monitoring analysis, d j is the jth performance evolution trend monitoring and analysis coefficient, and e is the eth item of data; The variable parameter relationship matrix D is established as: D=(g ij ) m×m×m In the formula, m×m×m is the combination of mine water level, water temperature and flow parameters; In the formula, g ij is the variable value of the combination of the i-th performance and the j-th performance, l ij is the evolution trend value of the combination of the i-th performance and the j-th performance, l ii is the evolution trend value of multiple i performance overlaps, l jj is the evolution trend value of multiple j performance overlaps, k i is the i-th performance evolution trend monitoring and analysis value, is the average value of the i-th performance evolution trend monitoring analysis, k j is the jth performance evolution trend monitoring and analysis value, The average value of the j-th performance evolution trend monitoring and analysis; Find the characteristic root and corresponding eigenvector of D, let the characteristic root a1≥a2…am≥0, and the corresponding eigenvector be c1≥c2…cm; Determine the principal components: the jth principal component W j The cumulative variance contribution rate of is: In the formula, Γ n is the variance contribution rate, n is the total number of j performances, m is one of the mine water level, water temperature and flow parameters, a j is the j performance variance contribution rate; Γ n When >70%, take the first m principal components W; Determine the entropy weight, calculate the comprehensive evolution trend monitoring and analysis value of dynamic monitoring accuracy, obtain the main component W = {W1, W2...Wm} and the corresponding characteristic root a = {a1≥a2...am}, obtain the comprehensive evolution trend monitoring and analysis value of dynamic monitoring accuracy, and judge the system performance.

10. A coal mine hydrological dynamic monitoring system based on neural network learning, characterized in that: The system is implemented by coal mine hydrological dynamic monitoring based on neural network learning by the method described in any one of claims 1 to 9, and the system comprises: A monitoring index acquisition module (1) for a coal mine hydrological dynamic evolution curve is used to perform coal mine hydrological dynamic image segmentation based on an improved neural network to acquire monitoring indexes for a coal mine hydrological dynamic evolution curve, wherein the coal mine hydrological dynamic evolution curve includes a mine water level, water temperature and flow data curve; Processing module (2), used for classifying and extracting the sources of monitoring index collection of coal mine hydrological dynamic evolution curve, performing vector processing of qualitative evolution trend monitoring and analysis, weighting of historical record timeliness, and normalization processing of original mine water level, water temperature and flow parameters; The monitoring and analysis module (3) is used for comprehensive evaluation of dynamic monitoring accuracy, performs standardization processing, determines the main components, determines the entropy weight, and makes judgments using the calculation formula of the comprehensive evolution trend monitoring and analysis value of dynamic monitoring accuracy.

Images