Method and device for predicting deformation of section coal pillar and electronic equipment

Abstract

The invention provides a deformation prediction method and device for a section coal pillar and electronic equipment. The method comprises the following steps: acquiring original monitoring data of a section coal pillar; determining target monitoring data according to the original monitoring data, an empirical mode decomposition algorithm and a differential evolution algorithm; according to the long short-term memory neural network and the back propagation neural network, an initial prediction model of the section coal pillar is constructed, and the initial prediction model has initial hyper-parameters; according to the initial prediction model and a genetic algorithm, target hyper-parameters are determined, a prediction training model is obtained, the genetic algorithm is used for performing optimization calculation on initial hyper-parameters of the initial prediction model, and the prediction training model is the initial prediction model with the target hyper-parameters; and determining target prediction data of the section coal pillar according to the target monitoring data and the prediction training model. According to the invention, the problem that the prediction method in the prior art is difficult to effectively predict the risk of rock burst is solved.

Description

Technical Field

[0001] The present application relates to the field of coal technology, and in particular to a method for predicting deformation of a section coal pillar, a device for predicting deformation of a section coal pillar, a computer-readable storage medium, and an electronic device. Background Art

[0002] With the steady improvement of coal resource mining methods, equipment and processes, large and medium-sized mines have gradually entered the stage of intelligent construction. At the same time, the increasing demand for coal resources has led to the depletion of shallow buried coal resources and coal resources with good geological conditions. As a result, coal resource mining has moved deep underground. Compared with the shallow underground environment, the deep underground environment shows new characteristics of "three highs and one disturbance", namely high stress, high temperature, high permeability pressure and strong mining disturbance. In particular, the rock burst (mine earthquake) disaster bred by the deep surrounding rock environment has attracted widespread attention in the field of underground coal mining.

[0003] Rock burst usually refers to the phenomenon of strong surrounding rock disturbance induced by deep underground space excavation or ore mining, which is mainly manifested by: sudden collapse of the surrounding rock and large deformation of the underground space floor. In severe cases, air waves or loud noises will occur, which will bring huge safety hazards to the deep underground coal mining process.

[0004] At present, the prediction of rock burst disasters is mainly carried out through theoretical analysis, indoor laboratory experiments and on-site monitoring. However, the occurrence of rock burst caused by deep underground coal mining is a very complex process, involving many influencing factors, including: the physical and mechanical properties of the coal body itself, the physical and mechanical properties of the overlying rock layers, underground mining process characteristics, tunnel support methods, ground stress effects and coal pillar size, etc. Therefore, it is difficult to effectively predict the danger of rock burst using traditional means and analysis methods.

[0005] Based on this, it is of great significance to establish an intelligent prediction method that takes into account multiple influencing factors for predicting the danger of rock burst during underground coal mining. Summary of the invention

[0006] The main purpose of the present application is to provide a method for predicting the deformation of a segmented coal pillar, a device for predicting the deformation of a segmented coal pillar, a computer-readable storage medium and an electronic device, so as to at least solve the problem that the prediction methods in the prior art are difficult to effectively predict the danger of rock burst.

[0007] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a deformation prediction method for a segmental coal pillar is provided, comprising: obtaining original monitoring data of the segmental coal pillar; determining target monitoring data based on the original monitoring data, an empirical mode decomposition algorithm and a differential evolution algorithm; constructing an initial prediction model of the segmental coal pillar based on a long short-term memory neural network and a back propagation neural network, the initial prediction model having initial hyperparameters; determining the target hyperparameters based on the initial prediction model and a genetic algorithm, and obtaining a prediction training model, the genetic algorithm being used to optimize the initial hyperparameters of the initial prediction model, the prediction training model being an initial prediction model with target hyperparameters; determining target prediction data of the segmental coal pillar based on the target monitoring data and the prediction training model.

[0008] Optionally, target monitoring data is determined according to the original monitoring data and the empirical mode decomposition algorithm and the differential evolution algorithm, including: determining the K-order modal component of the original monitoring data according to the original monitoring data and the empirical mode decomposition algorithm, the K-order modal component including multi-order components; determining the spread entropy value of the multi-order component according to the K-order modal component and the differential evolution algorithm; obtaining multiple target ratios according to the spread entropy values ​​of the multi-order components, and determining the component corresponding to the maximum ratio among the multiple target ratios as the target-order component, wherein the target ratio is the ratio of the spread entropy value of the k-th order component to the spread entropy value of the k+1-th order component, k≥1, and k is an integer; obtaining a reconstructed component, the reconstructed component being the component other than the two components corresponding to the maximum ratio in the multi-order components; and reconstructing the original monitoring data corresponding to the reconstructed component in the original monitoring data to obtain the target monitoring data.

[0009] Optionally, according to the original monitoring data and the empirical mode decomposition algorithm, the K-order modal component of the original monitoring data is determined, including: obtaining the first residual and the k+1th residual of the original monitoring data; determining the first-order component of the original monitoring data according to the first residual and a first preset formula, the first preset formula is: IMF1=x-R1, wherein IMF1 is the first-order component, x is the original monitoring data, and R1 is the first residual; determining the k+1th order component according to the first residual, the k+1th residual, and the second preset formula, the second preset formula is: IMF k+1 =R k -R k+1 , IMF k+1 is the k+1th order component, R k is the kth residual, R k+1 is the k+1th residual.

[0010] Optionally, obtaining the first residual and the k+1th residual of the original monitoring data includes: obtaining multiple original monitoring sequences, the multiple original monitoring sequences corresponding one-to-one to the multi-order components; determining the local mean of the first original monitoring sequence as the first residual, the first original monitoring sequence being obtained by adding white noise to the original monitoring data; obtaining the signal after adding white noise to the kth residual, and determining the local mean of the signal as the k+1th residual.

[0011] Optionally, according to the K-order modal components and the differential evolution algorithm, the scatter entropy values ​​of the multi-order components are determined, including: determining the target monitoring sequence according to the K-order modal components and the normal cumulative distribution function; determining multiple embedding vectors corresponding to the target monitoring sequence according to the target monitoring sequence and the embedding dimension; determining multiple probabilities of the K-order modal components according to the multiple embedding vectors and the differential evolution algorithm, and the multiple probabilities correspond one-to-one to the multi-order components; determining multiple scatter entropy values ​​corresponding one-to-one to the multi-order components according to the multiple probabilities and the scatter entropy definition.

[0012] Optionally, determining the target monitoring sequence according to the K-order modal component and the normal cumulative distribution function includes: determining a first monitoring sequence according to the K-order modal component and the normal cumulative distribution function; determining the target monitoring sequence according to the first monitoring sequence, a preset integer and a third preset formula, the third preset formula being: i =round(c*y i +0.5), where z i is the target monitoring sequence, y i is the first monitoring sequence, and c is a preset integer.

[0013] Optionally, the target monitoring data includes training set data and test set data, and the target prediction data of the section coal pillar is determined according to the target monitoring data and the prediction training model, including: determining a first prediction value according to the first historical data in the training set data and the prediction training model; judging whether the error between the first prediction value and the second historical data in the training set data is less than or equal to the target error, and obtaining a first judgment result; when the first judgment result indicates yes, stopping training the prediction training model, and determining the prediction training model as the target prediction model, and determining the target prediction data according to the test set data and the target prediction model.

[0014] According to another aspect of the present application, a deformation prediction device for a segmented coal pillar is provided, comprising: a data acquisition module for acquiring original monitoring data of the segmented coal pillar; a noise reduction module for determining target monitoring data based on the original monitoring data and an empirical mode decomposition algorithm and a differential evolution algorithm; a construction module for constructing an initial prediction model of the segmented coal pillar based on a long short-term memory neural network and a back propagation neural network, the initial prediction model having initial hyperparameters; an optimization module for determining target hyperparameters based on the initial hyperparameters and a genetic algorithm, and obtaining a prediction training model, the genetic algorithm being used to perform optimization calculations on the initial hyperparameters of the initial prediction model, the prediction training model being an initial prediction model with target hyperparameters; a prediction module for determining target prediction data of the segmented coal pillar based on the target monitoring data and the prediction training model.

[0015] According to another aspect of the present application, a computer-readable storage medium is provided, the computer-readable storage medium including a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the deformation prediction methods for the segmented coal pillar.

[0016] According to another aspect of the present application, an electronic device is provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include methods for executing any one of the deformation prediction methods for segmental coal pillars.

[0017] By applying the technical solution of the present application, the original monitoring data of the coal pillar of the section can be first obtained, and the initial prediction model can be constructed by the long short-term memory neural network and the back propagation neural network. Then, in order to improve the reliability of the original monitoring data, the technical solution of the present application uses the empirical mode decomposition algorithm and the differential evolution algorithm to perform noise reduction processing on the original monitoring data, and determines that the target monitoring data is obtained, and the target monitoring data is the original monitoring data after the noise reduction processing. And in order to improve the model performance of the initial prediction model, the technical solution of the present application also uses a genetic algorithm to optimize the initial hyperparameters of the initial prediction model, so that after the optimization calculation, the initial hyperparameters of the initial prediction model can be adjusted to the target hyperparameters to obtain a prediction training model. It can be understood that the prediction training model is an initial prediction model with target hyperparameters, so that the model performance of the prediction training model is better than the model performance of the initial prediction model with the initial hyperparameters. Finally, since the target monitoring data after noise reduction processing has a higher data set accuracy, the prediction training model with target hyperparameters has a higher model performance, so after the target monitoring data is input into the prediction training model, prediction data with higher prediction accuracy can be obtained, that is, the accuracy of the target prediction data of the section coal pillar is determined to be high, which has a good effect in predicting the deformation of the section coal pillar. Therefore, the technical solution of this application solves the problem that the prediction method in the prior art is difficult to effectively predict the danger of rock burst. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings constituting part of the present application are used to provide a further understanding of the present application. The exemplary embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0019] Figure 1 A hardware structure block diagram of a mobile terminal for executing a deformation prediction method of a section coal pillar provided in an embodiment of the present application is shown;

[0020] Figure 2 A schematic flow chart of a method for predicting deformation of a segmented coal pillar provided according to an embodiment of the present application is shown;

[0021] Figure 3 A structural block diagram of a device for predicting deformation of a segmented coal pillar provided according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0022] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0023] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0024] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0025] As introduced in the background technology, coal resource mining is moving towards deep underground. Compared with the shallow underground environment, the deep underground environment shows new characteristics of "three highs and one disturbance", namely high stress, high temperature, high permeability pressure and strong mining disturbance. In particular, the rock burst (mine earthquake) disaster bred by the deep surrounding rock environment has attracted extensive attention in the field of underground coal mining. In the prior art, the prediction of rock burst disasters is mainly carried out through theoretical analysis, indoor laboratory experiments and on-site monitoring. However, the occurrence of rock burst caused by deep underground coal mining is a very complicated process, which involves many influencing factors, including: the physical and mechanical properties of the coal body itself, the physical and mechanical properties of the overlying rock formations, the characteristics of the underground mining process, the tunnel support method, the ground stress effect and the size of the coal pillar. Therefore, it is difficult to effectively predict the danger of rock burst using traditional means and analysis methods. In order to solve the problem that the prediction method in the prior art is difficult to effectively predict the danger of rock burst, the embodiment of the present application provides a deformation prediction method of a section coal pillar, a deformation prediction device of a section coal pillar, a computer-readable storage medium and an electronic device.

[0026] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0027] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 FIG. 1 is a hardware structure block diagram of a mobile terminal for predicting deformation of a segmented coal pillar according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0028] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the deformation prediction method of the section coal pillar in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0029] In this embodiment, a method for predicting deformation of a segmented coal pillar that runs on a mobile terminal, a computer terminal, or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0030] Figure 2FIG. 1 is a flow chart of a method for predicting deformation of a segmented coal pillar according to an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0031] Step S201, obtaining original monitoring data of the section coal pillar;

[0032] Specifically, the above-mentioned original monitoring data can be fiber Bragg grating strain monitoring data, and the original monitoring data can be obtained by studying the influence range of the overlying flow oblique coal pillar and the deformation law of the section coal pillar along the strike and width directions. Optionally, a section coal pillar deformation sample data set can be constructed through the above-mentioned original monitoring data to be used for intelligent prediction of the deformation of the section coal pillar.

[0033] Step S202, determining target monitoring data according to the original monitoring data, the empirical mode decomposition algorithm and the differential evolution algorithm;

[0034] Specifically, this step reduces the noise of the original monitoring data in step S201 by introducing a denoising algorithm that combines an empirical mode decomposition algorithm (ICEEMDAN) and a differential evolution algorithm (DE), removes redundant data, and thereby obtains the true data of the deformation of the coal pillar in the section, i.e., the above-mentioned target monitoring data.

[0035] Step S203, constructing an initial prediction model of the section coal pillar according to the long short-term memory neural network and the back propagation neural network, wherein the initial prediction model has initial hyperparameters;

[0036] Specifically, the initial prediction model can predict the deformation of the coal pillar in the section corresponding to the fiber Bragg grating strain monitoring data. It can be understood that the initial hyperparameters define the model structure of the initial prediction model and control the learning process of the initial prediction model. In this step, the initial prediction model constructed by combining the long short-term memory neural network and the back propagation neural network takes into account the complexity of the working environment and the lack of obvious regularity of the original monitoring data. Therefore, the use of this initial prediction model for prediction can improve the accuracy of the prediction.

[0037] Optionally, the initial prediction model can be an LSTM-BP neural network model. In the initial prediction model, the long short-term memory neural network (LSTM) uses three gate units to control the transmission of information, namely the forget gate, the input gate and the output gate. The back propagation neural network (BP) in the initial prediction model is a neural network that calculates the loss function and propagates the gradient information from the output layer to the input layer, and is mainly composed of an input layer, a hidden layer and an output layer.

[0038] Specifically, the specific calculation formula of the LSTM neural network is:

[0039] Forget gate: y t =σ(Wxf x t +W hf h t-1 +b f );

[0040] z i =σ(W xi x t +W hi h t-1 +b i )

[0041] Input gate: z = tanh(W xc x t +W hc h t-1 +b c );

[0042] c t =y t ·c t-1 +z i ·z

[0043] Output Gate:

[0044] The input gate is used to update the node state, and the input is the current moment information and the hidden state of the previous moment; the forget gate selectively forgets the information to prevent minor information from affecting the prediction results; the hidden state output by the output gate contains the information input at the previous moment, and the hidden state and the input at the current moment are further transmitted to the input gate of the next moment unit to complete the information transmission, so that the model has a memory function. · represents element-by-element multiplication, W xi , W xc and W xf Represents input data x in different node states t The weight parameter, W hi , W hc and W hf Indicates the hidden state h in different node states t-1 The weight parameter, b c 、b f 、b i Represents the bias under different unit states.

[0045] The specific calculation formula of BP neural network is:

[0046] Input layer: x i ;

[0047] Hidden Layer:

[0048] Output layer:

[0049] The process of BP neural network is mainly divided into two stages. The first stage is the forward propagation of the signal, from the input layer through the hidden layer, and finally to the output layer; the second stage is the reverse propagation of the error, from the output layer to the hidden layer, and finally to the input layer, adjusting the weights and biases from the hidden layer to the output layer, and the weights and biases from the input layer to the hidden layer in turn. i is the h-th layer input, w ih is the i-th input node, α i is the current node value; y i Represents the final output value of the neural network.

[0050] Step S204, according to the initial prediction model and the genetic algorithm, the target hyperparameters are determined, and a prediction training model is obtained. The genetic algorithm is used to optimize the initial hyperparameters of the initial prediction model. The prediction training model is the initial prediction model with the target hyperparameters.

[0051] Specifically, before model training, in order to improve the model performance of the initial prediction model, a genetic algorithm can be used to optimize the initial hyperparameters of the initial prediction model, so as to obtain the optimal hyperparameter combination of the initial prediction model, that is, the target hyperparameter. It can be understood that when the target hyperparameter is found, the target hyperparameter can be applied to the initial prediction model to obtain a prediction training model, which is the initial prediction model with the target hyperparameter.

[0052] Optionally, a genetic algorithm (GA) is used to optimize the initial hyperparameters of the initial prediction model (LSTM-BP neural network model). The specific calculation process is shown in the following steps:

[0053] S2041, taking the initial hyperparameters such as the number of hidden layers and weights of the initial prediction model (LSTM-BP neural network model) as the problem to be optimized, and randomly generating n chromosomes of a certain length consistent with the problem to be optimized as the initial population;

[0054] S2042, calculate the fitness value;

[0055] S2043, by applying the selection operator to the group, excellent individuals are selected according to their individual fitness values;

[0056] S2044, through the crossover operator acting on the population, the selected pairs of individuals exchange part of the chromosomes with a certain probability to generate new individuals;

[0057] S2045, through the mutation operator acting on the population, the gene value of the selected individual is changed with a certain probability;

[0058] S2046, compare the fitness value of the new individual in S2045 with the required fitness threshold. If it is less than the threshold, stop the loop and obtain the weight and threshold at this time. If it is greater than the threshold, perform the next genetic operation.

[0059] Step S205, determining target prediction data of the section coal pillar according to the target monitoring data and the prediction training model.

[0060] Specifically, it is usually necessary to train the prediction training model to improve the model reliability of the prediction training model. Based on this, this step first trains the prediction training model using the target monitoring data that has been obtained, and then when the prediction training model reaches the condition for stopping training, the target monitoring data can be input into the above prediction training model again to obtain the target prediction data of the section coal pillar. It can be understood that the data accuracy of the target prediction data is relatively high, and it can be used to predict the deformation of the section coal pillar, which can solve the problem that the prediction method in the prior art is difficult to effectively predict the danger of rock burst.

[0061] It is worth mentioning here that the deformation and damage of the segmental coal pillar can be used as a predictive indicator of rock burst danger. By monitoring the deformation of the segmental coal pillar, the stability of the mine and the risk of rock burst can be evaluated.

[0062] Through the above-mentioned embodiments, the original monitoring data of the coal pillar of the section can be first obtained, and the initial prediction model can be constructed by the long short-term memory neural network and the back propagation neural network. Afterwards, in order to improve the reliability of the original monitoring data, the technical solution of the present application uses the empirical mode decomposition algorithm and the differential evolution algorithm to perform noise reduction processing on the original monitoring data, and determines that the target monitoring data is obtained, and the target monitoring data is the original monitoring data after the noise reduction processing. And in order to improve the model performance of the initial prediction model, the technical solution of the present application also uses a genetic algorithm to perform an optimization calculation on the initial hyperparameters of the initial prediction model, so that after the optimization calculation, the initial hyperparameters of the initial prediction model can be adjusted to the target hyperparameters to obtain a prediction training model. It can be understood that the prediction training model is an initial prediction model with target hyperparameters, so that the model performance of the prediction training model is better than the model performance of the initial prediction model with the initial hyperparameters. Finally, since the target monitoring data after noise reduction processing has a higher data set accuracy, the prediction training model with target hyperparameters has a higher model performance, so after the target monitoring data is input into the prediction training model, prediction data with higher prediction accuracy can be obtained, that is, the accuracy of the target prediction data of the section coal pillar is determined to be high, which has a good effect in predicting the deformation of the section coal pillar. Therefore, the technical solution of this application solves the problem that the prediction method in the prior art is difficult to effectively predict the danger of rock burst.

[0063] In order to remove noise in the original monitoring data, the original monitoring data can be subjected to denoising processing, that is, in some optional implementations, the target monitoring data is determined according to the original monitoring data and the empirical mode decomposition algorithm and the differential evolution algorithm, including: determining the K-order modal component of the original monitoring data according to the original monitoring data and the empirical mode decomposition algorithm, the K-order modal component including multi-order components; determining the spread entropy value of the multi-order component according to the K-order modal component and the differential evolution algorithm; obtaining multiple target ratios according to the spread entropy value of the multi-order components, and determining the component corresponding to the maximum ratio among the multiple target ratios as the target-order component, wherein the target ratio is the ratio of the spread entropy value of the k-th order component to the spread entropy value of the k+1-th order component, k≥1, and k is an integer; obtaining a reconstructed component, the reconstructed component being the component other than the two components corresponding to the maximum ratio in the multi-order components; reconstructing the original monitoring data corresponding to the reconstructed component in the original monitoring data to obtain the target monitoring data.

[0064] Specifically, the empirical mode decomposition algorithm is used to decompose the above-mentioned original monitoring data into K-order modal components. Among them, the K-order modal components correspond to K different modal components, which are the manifestations of the signal corresponding to the original monitoring data at different frequencies, and each modal component corresponds to a specific frequency component in the signal. Exemplarily, the K modal components include multi-order components, and the multi-order components are respectively the 1st order component, the 2nd order component, ..., the k-1th order component, the kth order component and the k+1th order component.

[0065] Specifically, the differential evolution algorithm can be used to optimize the calculation process of the spread entropy value of the K-order modal component. In other words, the differential evolution algorithm can perform parameter optimization and feature selection when calculating the spread entropy value, improve the efficiency of the spread entropy calculation, and make the spread entropy value of the determined multi-order component more accurate.

[0066] It is clear that the larger the spread entropy value is, the higher the complexity of the original monitoring sequence is; conversely, the smaller the value is, the lower its complexity is.

[0067] Specifically, a spread entropy value is calculated for each order component, so the number of spread entropy values ​​of multi-order components is multiple.

[0068] On this basis, by calculating the ratio of two spread entropy values ​​corresponding to adjacent order components, multiple spread entropy ratios can be obtained, and the above target ratio can be the maximum ratio among the multiple spread entropy ratios. Exemplarily, the ratio of two spread entropy values ​​corresponding to adjacent order components can include: the ratio of the spread entropy value of the 1st order component to the spread entropy value of the 2nd order component, recorded as the first spread entropy ratio; the ratio of the spread entropy value of the 2nd order component to the spread entropy value of the 3rd order component, recorded as the second spread entropy ratio; ...; the ratio of the spread entropy value of the k-1th order component to the spread entropy value of the kth order component, recorded as the k-1th spread entropy ratio; the ratio of the spread entropy value of the kth order component to the spread entropy value of the k+1th order component, recorded as the kth spread entropy ratio, then the above target ratio can be the largest spread entropy ratio among the first spread entropy ratio, the second spread entropy ratio, ..., the k-1th spread entropy ratio and the kth spread entropy ratio.

[0069] Optionally, the noise component can be removed with the above target ratio as a threshold limit, that is, the two components corresponding to the target ratio can be used as noise components, so that the two components corresponding to the target ratio can be removed from the K-stage modal component, and the remaining components in the multi-order components of the K-stage modal component can be used as reconstructed components. Furthermore, the original monitoring data corresponding to the remaining components in the K-stage modal component can be reconstructed to obtain the target monitoring data. Exemplarily, if the largest spread entropy ratio among the first spread entropy ratio, the second spread entropy ratio, ..., the k-1th spread entropy ratio and the kth spread entropy ratio is the kth spread entropy ratio, then the kth-order component and the k+1th-order component are noise components.

[0070] Exemplarily, for IMF1, IMF2, IMF3, IMF4, IMF5, IMF6, IMF7 and IMF8, if the ratio of IMF2 to IMF1 is the maximum ratio, since the ratio of IMF1 and IMF2 is the largest, it can be determined that IMF1 and IMF2 are noise components, and IMF3, IMF4, IMF5, IMF6, IMF7 and IMF8 are effective components, so the original monitoring data corresponding to IMF3, IMF4, IMF5, IMF6, IMF7 and IMF8 can be added and reconstructed to obtain target monitoring data. At this time, the target monitoring data is the monitoring data after noise reduction processing, which can be used as training data for the prediction training model and prediction data corresponding to the target prediction data.

[0071] In order to obtain the K-order modal component of the original monitoring data, in some optional implementations, the K-order modal component of the original monitoring data is determined according to the original monitoring data and the empirical mode decomposition algorithm, including: obtaining the first residual and the k+1th residual of the original monitoring data; determining the first-order component of the original monitoring data according to the first residual and the first preset formula, the first preset formula is: IMF1=x-R1, wherein IMF1 is the first-order component, x is the original monitoring data, and R1 is the first residual; determining the k+1th order component according to the first residual, the k+1th residual, and the second preset formula, the second preset formula is: IMF k+1 =R k -R k+1 , IMF k+1 is the k+1th order component, R k is the kth residual, R k+1 is the k+1th residual.

[0072] Among them, the step of obtaining the first residual and the k+1th residual of the original monitoring data may include: obtaining multiple original monitoring sequences, the multiple original monitoring sequences correspond one-to-one to the multi-order components; determining that the local mean of the first original monitoring sequence is the first residual, and the first original monitoring sequence is obtained by adding white noise to the original monitoring data; obtaining the signal after adding white noise to the kth residual, and determining that the local mean of the signal is the k+1th residual.

[0073] Specifically, the above embodiment constructs an original monitoring sequence by adding white noise to the original monitoring data.

[0074] For example, the original monitoring sequence x can be constructed by the following formula: (i) :

[0075] x (i) =x+β0E[ω (i) ], x is the original monitoring data x (i) is the original monitoring sequence, β0 is the standard deviation of white noise, and Where E is the K-order modal component generated by EMD decomposition, and std is the standard deviation operator.

[0076] Specifically, the first residual R1=N(x (i) ), where N represents the local mean of the generated signal.

[0077] It is understandable that in order to obtain the k+1th residual R k+1 , we can first obtain the signal after adding white noise to the kth residual. Based on the original monitoring sequence x (i) This principle is obtained by adding white noise to the original monitoring data. The signal after adding white noise to the kth residual can be R k +βk E[ω (i) ], so the k+1th residual R k+1 Can be: R k+1 =N(R k +β k E[ω (i) ]), where β k is the standard deviation of the white noise superimposed on the original monitoring data.

[0078] For example, when k=1, the second residual R2 can be expressed as R2=N(R1+β1E[ω (i) ]).

[0079] Exemplarily, in order to verify the effectiveness of the empirical mode decomposition algorithm and differential evolution algorithm (ICEEMDAN-DE) proposed in the embodiment of the present application, different noise reduction methods are used to reduce the noise of the original monitoring data of monitoring point No. 8 in hole No. 3, and the comparison results (mean square error (RMSE)) are shown in Table 1:

[0080] Table 1

[0081]

[0082] From the comparison of the denoising performance indicators of the original monitoring data using different algorithms, it can be seen that the mean square error (RMSE) based on the ICEEMDAN-DE denoising model is reduced by an average of 1.02 compared with the mean square error (RMSE) based on the CEEMDAN denoising method; the mean square error (RMSE) based on the ICEEMDAN-DE denoising model is reduced by an average of 2.33 compared with the mean square error (RMSE) based on the EEMD denoising algorithm.

[0083] Comprehensive comparison shows that the ICEEMDAN-DE denoising model further improves the signal-to-noise ratio, reduces the mean square error, and has a better denoising effect than the CEEMDAN and EEMD denoising models.

[0084] In some optional embodiments, the scatter entropy values ​​of the multi-order components are determined according to the K-order modal components and the differential evolution algorithm, including: determining the target monitoring sequence according to the K-order modal components and the normal cumulative distribution function; determining multiple embedding vectors corresponding to the target monitoring sequence according to the target monitoring sequence and the embedding dimension; determining multiple probabilities of the K-order modal components according to the multiple embedding vectors and the differential evolution algorithm, and the multiple probabilities correspond one-to-one to the multi-order components; determining multiple scatter entropy values ​​corresponding one-to-one to the multi-order components according to the multiple probabilities and the scatter entropy definition.

[0085] Wherein, in order to obtain the target monitoring sequence, the target monitoring sequence is determined according to the K-order modal component and the normal cumulative distribution function, including: determining the first monitoring sequence according to the K-order modal component and the normal cumulative distribution function; determining the target monitoring sequence according to the first monitoring sequence, a preset integer and a third preset formula, and the third preset formula is: i =round(c*y i +0.5), where z i is the target monitoring sequence, y i is the first monitoring sequence, c is a preset integer, and round means rounding to the nearest integer.

[0086] Specifically, for each order component in the K-order modal component, the normal cumulative distribution function can be used to integrate the order component to obtain the first monitoring sequence of the order component. Further, a preset integer can be selected to round the first monitoring sequence to obtain a target monitoring sequence.

[0087] Optionally, the first monitoring sequence y obtained by integrating the components of the K-order modal components using the normal cumulative distribution function is j The formula is:

[0088] Among them, X i is the data point in the component of the K-order modal component, σ is the standard deviation, and u is the mean.

[0089] Exemplarily, for the first-order component IMF1, if IMF1=[9, 8, 1, 12, 5, -3, 1.5], the first monitoring sequence corresponding to IMF1 may be: y=[0.82, 0.75, 0.21, 0.94, 0.52, 0.05, 0.241].

[0090] Exemplarily, for the first monitoring sequence of the first-order component IMF1, the preset integer 3 can be selected, so that through the above third preset formula, the target monitoring sequence can be obtained as: z=[3, 3, 1, 3, 2, 1, 1].

[0091] Specifically, the embedding vector is created by reorganizing the corresponding target monitoring sequence of the modal component into a two-dimensional matrix, each row in the two-dimensional matrix represents an embedding vector, and each embedding vector is composed of two adjacent data points in the target monitoring sequence.

[0092] For example, for each modal component IMF x For this modal component IMF x The corresponding multiple embedding vectors can be Where m is the embedding dimension, c is the number of categories, d is the time delay, i = 1, 2, ..., N-(m-1)d, so the two-dimensional matrix corresponding to the modal component is

[0093] Specifically, multiple scatter patterns can be determined based on the clustering algorithm and multiple embedding vectors. The scatter pattern is a subsequence that appears repeatedly in the target monitoring sequence. For example, in the target monitoring sequence z = [3, 3, 1, 3, 2, 1, 1], if the embedding dimension m = 2, then since [3, 1] in the target monitoring sequence appears repeatedly, [3, 1] is a scatter pattern of the target monitoring sequence.

[0094] For example, if but The corresponding scattering pattern is πv0v1...v m-1 ,v=1,2,...,c. Each scatter mode consists of c data points, and each data point has m values, so the modal component IMF x The corresponding scattering pattern is c m kind.

[0095] Specifically, the differential evolution algorithm can be used to find the best scattering pattern, so that according to multiple embedding vectors and the differential evolution algorithm, the best scattering pattern can be determined from multiple scattering patterns. It can be understood that each modal component can correspond to a target scattering pattern, that is, the best scattering pattern among multiple scattering patterns, so that when the K-order modal component includes multiple modal components, the K-order modal component can correspond to multiple target scattering patterns, and each target scattering pattern corresponds to a modal component in the K-order modal component.

[0096] Specifically, for each modal component IMF x The above target dispersion pattern can be used to calculate the number of times the subsequence corresponding to the target dispersion pattern appears in all subsequences of the target monitoring sequence, so as to obtain the corresponding modal component IMF x The probability of the target dispersion mode. It can be understood that each modal component corresponds to a probability, so when the K-order modal component includes multiple modal components, multiple probabilities of the K-order modal component can be obtained, and the multiple probabilities correspond to the multiple modal components (i.e., multi-order components) one by one. Exemplarily, the probability of each target dispersion mode can be calculated according to the following formula:

[0097]

[0098] in, Represents the embedding vector The number mapped to the scatter entropy pattern is the number of times the subsequence corresponding to the target scatter pattern appears in all subsequences of the target monitoring sequence.

[0099] Further, in the step of determining multiple scatter entropy values ​​corresponding to the multi-order components one by one according to the multiple probabilities and scatter entropy definitions, the scatter entropy DE(x,m,c,d) can be calculated by the following formula:

[0100]

[0101] In some optional embodiments, the target monitoring data includes training set data and test set data, and the target prediction data of the section coal pillar is determined according to the target monitoring data and the prediction training model, including: determining a first prediction value according to the first historical data in the training set data and the prediction training model; judging whether the error between the first prediction value and the second historical data in the training set data is less than or equal to the target error, and obtaining a first judgment result; if the first judgment result indicates yes, stopping training the prediction training model, and determining the prediction training model as the target prediction model, and determining the target prediction data according to the test set data and the target prediction model.

[0102] Optionally, 80% of the target monitoring data can be used as training set data, and 20% can be used as test set data.

[0103] Specifically, the above-mentioned target monitoring data can be historical data, so that the above-mentioned training set data and test set data can also be recorded as the first historical data. It is mentioned here that the first historical data and the second historical data can be used as sample values ​​for training the above-mentioned prediction training model, that is, the second historical data can be used as the ideal situation corresponding to the prediction value of the first historical data, so that the embodiment of the present application can determine the first prediction value according to the first historical data in the above-mentioned training set data and the prediction training model, and can judge whether the prediction training model is reliable by comparing the error between the first prediction value and the second historical data.

[0104] It can be understood that when the above-mentioned first judgment result indicates no, the prediction training model can continue to be trained according to the above-mentioned training set data until the prediction training model, when the first historical data is input, the error between the first prediction value output by the prediction training model and the second historical data is less than or equal to the target error.

[0105] In addition, in some optional implementations, the target prediction data can be evaluated by four indicators. Optionally, the four indicators can be absolute error (MAE), relative error (MSE), root mean square error (RMSE) and linear regression correlation coefficient (R 2 ). The smaller the MAE, MSE and RMSE values, the better. 2The numerical value range is between 0 and 1. The closer it is to 1, the better the curve fitting effect is. The numerical results of the indicators show the fitting of the predicted curve and the true curve, thus reflecting the prediction accuracy.

[0106] For example, in order to further verify the performance of the segmental coal pillar deformation prediction model based on the empirical mode decomposition algorithm-differential evolution algorithm-genetic algorithm-long short-term memory neural network-back propagation neural network (ICEEMDAN-DE-GA-LSTM-BP), the prediction results corresponding to the data of three measuring holes (hole 1, hole 3 and hole 4) of the 5-1# segment coal pillar were selected for comparative analysis. Among them, the prediction results corresponding to the data of hole 1 are shown in Table 2, the prediction results corresponding to hole 3 are shown in Table 3, and the prediction results corresponding to hole 4 are shown in Table 4:

[0107] Table 2

[0108]

[0109] Table 3

[0110]

[0111] Table 4

[0112]

[0113] In Table 2, when the working face advances to 1360.5m, the first small peak value of strain at monitoring point 5 near the mining side of the 22208 working face appears. At this time, the actual peak value is 304.39, and the predicted peak value is 313.01. The predicted MAE, MSE, RMSE, and R 2 They are 6.93, 55.31, 7.44, and 0.89 respectively. Among them, MAE, MSE, and RMSE are relatively small and R 2 The value can reach 0.89, which is close to 1, and the accuracy is about 89%.

[0114] When the working face advances to 1382.5 m in Table 3, it is already in the stage of passing the remaining oblique coal pillar. The strain change curves and strain trend prediction curves of the seven monitoring points are shown in Table 3. 2 Both are around 0.95, and the accuracy is about 95%.

[0115] As shown in Table 4, as the working face advances to 1436.5m, it is already in the stage of the residual oblique coal pillar, and the strain curve changes with an upward trend, but the deformation of the coal pillar in this area is small and is in a relatively stable range. The actual strain curve is slightly different from the predicted curve. According to the evaluation index, the R value of the strain trend prediction curve of each monitoring point is 2 Both are around 0.96, with an accuracy of about 96%.

[0116] Compared with the prior art, the prediction method of the embodiment of the present application can not only monitor the deformation of the segmented coal pillar in real time, but also realize the intelligent prediction of the deformation of the segmented coal pillar under the legacy oblique coal pillar. Combining the neural network technology with the prediction of segmented coal pillar deformation lays a foundation for combining coal pillar deformation with artificial intelligence to further expand the application scope of neural network technology. In addition, the entire model training process in the embodiment of the present application does not require human subjective participation in the analysis task. The accuracy of the deep learning model is higher than that of a single model, which has important guiding significance for the prediction of the deformation of the segmented coal pillar under the legacy coal pillar.

[0117] The embodiment of the present application also provides a deformation prediction device for a section coal pillar. It should be noted that the deformation prediction device for a section coal pillar in the embodiment of the present application can be used to execute the deformation prediction method for a section coal pillar provided in the embodiment of the present application. The device is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0118] The following is an introduction to the deformation prediction device for a segmented coal pillar provided in an embodiment of the present application.

[0119] Figure 3 Schematic diagram of a deformation prediction device for a segmented coal pillar according to an embodiment of the present application. Figure 3 As shown, the device comprises:

[0120] The data acquisition module 10 is used to obtain the original monitoring data of the section coal pillar;

[0121] Specifically, the above-mentioned original monitoring data can be fiber Bragg grating strain monitoring data, and the original monitoring data can be obtained by studying the influence range of the overlying flow oblique coal pillar and the deformation law of the section coal pillar along the strike and width directions. Optionally, a section coal pillar deformation sample data set can be constructed through the above-mentioned original monitoring data to be used for intelligent prediction of the deformation of the section coal pillar.

[0122] The noise reduction module 20 is used to determine the target monitoring data according to the original monitoring data, the empirical mode decomposition algorithm and the differential evolution algorithm;

[0123] Specifically, this step reduces the noise of the original monitoring data in step S201 by introducing a denoising algorithm that combines an empirical mode decomposition algorithm (ICEEMDAN) and a differential evolution algorithm (DE), removes redundant data, and thereby obtains the true data of the deformation of the coal pillar in the section, i.e., the above-mentioned target monitoring data.

[0124] A construction module 30 is used to construct an initial prediction model of the section coal pillar according to the long short-term memory neural network and the back propagation neural network, wherein the initial prediction model has initial hyperparameters;

[0125] Specifically, the initial prediction model can predict the deformation of the coal pillar in the section corresponding to the fiber Bragg grating strain monitoring data. It can be understood that the initial hyperparameters define the model structure of the initial prediction model and control the learning process of the initial prediction model. In this step, the initial prediction model constructed by combining the long short-term memory neural network and the back propagation neural network takes into account the complexity of the working environment and the lack of obvious regularity of the original monitoring data. Therefore, the use of this initial prediction model for prediction can improve the accuracy of the prediction.

[0126] Optionally, the initial prediction model can be an LSTM-BP neural network model. In the initial prediction model, the long short-term memory neural network (LSTM) uses three gate units to control the transmission of information, namely the forget gate, the input gate and the output gate. The back propagation neural network (BP) in the initial prediction model is a neural network that calculates the loss function and propagates the gradient information from the output layer to the input layer, and is mainly composed of an input layer, a hidden layer and an output layer.

[0127] Specifically, the specific calculation formula of the LSTM neural network is:

[0128] Forget gate: y t =σ(W xf x t +W hf h t-1 +b f );

[0129] z i =σ(W xi x t +W hi h t-1 +b i )

[0130] Input gate: z = tanh(W xc x t +W hc h t-1 +b c );

[0131] c t =y t ·c t-1 +z i ·z

[0132] Output Gate:

[0133] The input gate is used to update the node state, and the input is the current moment information and the hidden state of the previous moment; the forget gate selectively forgets the information to prevent minor information from affecting the prediction results; the hidden state output by the output gate contains the information input at the previous moment, and the hidden state and the input at the current moment are further transmitted to the input gate of the next moment unit to complete the information transmission, so that the model has a memory function. · represents element-by-element multiplication, W xi , W xc and W xf Represents input data x in different node states t The weight parameter, W hi , W hc and W hf Indicates the hidden state h in different node states t-1 The weight parameter, b c 、b f 、b i Represents the bias under different unit states.

[0134] The specific calculation formula of BP neural network is:

[0135] Input layer: x i ;

[0136] Hidden Layer:

[0137] Output layer:

[0138] The process of BP neural network is mainly divided into two stages. The first stage is the forward propagation of the signal, from the input layer through the hidden layer, and finally to the output layer; the second stage is the reverse propagation of the error, from the output layer to the hidden layer, and finally to the input layer, adjusting the weights and biases from the hidden layer to the output layer, and the weights and biases from the input layer to the hidden layer in turn. i is the h-th layer input, w ih is the i-th input node, α i is the current node value; y i Represents the final output value of the neural network.

[0139] The optimization module 40 is used to determine the target hyperparameters and obtain the prediction training model according to the initial prediction model and the genetic algorithm. The genetic algorithm is used to optimize the initial hyperparameters of the initial prediction model. The prediction training model is the initial prediction model with the target hyperparameters.

[0140] Specifically, before model training, in order to improve the model performance of the initial prediction model, a genetic algorithm can be used to optimize the initial hyperparameters of the initial prediction model, so as to obtain the optimal hyperparameter combination of the initial prediction model, that is, the target hyperparameter. It can be understood that when the target hyperparameter is found, the target hyperparameter can be applied to the initial prediction model to obtain a prediction training model, which is the initial prediction model with the target hyperparameter.

[0141] Optionally, a genetic algorithm (GA) is used to optimize the initial hyperparameters of the initial prediction model (LSTM-BP neural network model). The specific calculation process is shown in the following steps:

[0142] S2041, taking the initial hyperparameters such as the number of hidden layers and weights of the initial prediction model (LSTM-BP neural network model) as the problem to be optimized, and randomly generating n chromosomes of a certain length consistent with the problem to be optimized as the initial population;

[0143] S2042, calculate the fitness value;

[0144] S2043, by applying the selection operator to the group, excellent individuals are selected according to their individual fitness values;

[0145] S2044, through the crossover operator acting on the population, the selected pairs of individuals exchange part of the chromosomes with a certain probability to generate new individuals;

[0146] S2045, through the mutation operator acting on the population, the gene value of the selected individual is changed with a certain probability;

[0147] S2046, compare the fitness value of the new individual in S2045 with the required fitness threshold. If it is less than the threshold, stop the loop and obtain the weight and threshold at this time. If it is greater than the threshold, perform the next genetic operation.

[0148] The prediction module 50 is used to determine the target prediction data of the section coal pillar according to the target monitoring data and the prediction training model.

[0149] Specifically, it is usually necessary to train the prediction training model to improve the model reliability of the prediction training model. Based on this, this step first trains the prediction training model using the target monitoring data that has been obtained, and then when the prediction training model reaches the condition for stopping training, the target monitoring data can be input into the above prediction training model again to obtain the target prediction data of the section coal pillar. It can be understood that the data accuracy of the target prediction data is relatively high, and it can be used to predict the deformation of the section coal pillar, which can solve the problem that the prediction method in the prior art is difficult to effectively predict the danger of rock burst.

[0150] It is worth mentioning here that the deformation and damage of the section coal pillar can be used as a predictive indicator of rock burst danger. By monitoring the deformation of the section coal pillar, the stability of the mine and the risk of rock burst can be evaluated.

[0151] Optionally, the noise reduction module includes: a first determination unit, used to determine the K-order modal component of the original monitoring data based on the original monitoring data and the empirical mode decomposition algorithm, the K-order modal component including multi-order components; a second determination unit, used to determine the spread entropy value of the multi-order component based on the K-order modal component and the differential evolution algorithm; a third determination unit, used to obtain multiple target ratios based on the spread entropy value of the multi-order component, and determine the component corresponding to the maximum ratio among the multiple target ratios as the target order component, wherein the target ratio is the ratio of the spread entropy value of the k-th order component to the spread entropy value of the k+1-th order component, k≥1, k is an integer; a first acquisition unit, used to obtain a reconstructed component, the reconstructed component is a component other than the two components corresponding to the maximum ratio in the multi-order components; a reconstruction unit, used to reconstruct the original monitoring data corresponding to the reconstructed component in the original monitoring data to obtain the target monitoring data.

[0152] Optionally, the first determination unit includes: a second acquisition unit, used to acquire a first residual and a k+1th residual of the original monitoring data; a first subunit, used to determine a first-order component of the original monitoring data according to the first residual and a first preset formula, the first preset formula being: IMF1=x-R1, wherein IMF1 is a first-order component, x is the original monitoring data, and R1 is a first residual; a second subunit, used to determine a k+1th order component according to the first residual, the k+1th residual, and a second preset formula, the second preset formula being: IMF k+1 =R k -R k+1 , IMF k+1 is the k+1th order component, R k is the kth residual, R k+1 is the k+1th residual.

[0153] Optionally, the second acquisition unit includes: a third subunit, used to acquire multiple original monitoring sequences, the multiple original monitoring sequences correspond one-to-one to the multi-order components; a fourth subunit, used to determine the local mean of the first original monitoring sequence as the first residual, the first original monitoring sequence is obtained by adding white noise to the original monitoring data; a fifth subunit, used to acquire the signal after adding white noise to the kth residual, and determine the local mean of the signal as the k+1th residual.

[0154] Optionally, the second determination unit includes: a sixth subunit, used to determine the target monitoring sequence according to the K-order modal components and the normal cumulative distribution function; a seventh subunit, used to determine multiple embedding vectors corresponding to the target monitoring sequence according to the target monitoring sequence and the embedding dimension; an eighth subunit, used to determine multiple probabilities of the K-order modal components according to the multiple embedding vectors and the differential evolution algorithm, and the multiple probabilities correspond one-to-one to the multi-order components; a ninth subunit, used to determine multiple scatter entropy values ​​corresponding one-to-one to the multi-order components according to the multiple probabilities and the scatter entropy definition.

[0155] Optionally, the sixth subunit includes: a first calculation subunit, used to determine the first monitoring sequence according to the K-order modal component and the normal cumulative distribution function; a second calculation subunit, used to determine the target monitoring sequence according to the first monitoring sequence, a preset integer and a third preset formula, the third preset formula is: i =round(c*y i +0.5), where z i is the target monitoring sequence, y i is the first monitoring sequence, and c is a preset integer.

[0156] Optionally, the target monitoring data includes training set data and test set data, and the prediction module includes: a fourth determination unit, used to determine a first prediction value based on the first historical data in the training set data and the prediction training model; a judgment unit, used to judge whether the error between the first prediction value and the second historical data in the training set data is less than or equal to the target error, and obtain a first judgment result; a prediction unit, used to stop training the prediction training model when the first judgment result indicates yes, and determine the prediction training model as the target prediction model, and determine the target prediction data based on the test set data and the target prediction model.

[0157] The deformation prediction device of the section coal pillar includes a processor and a memory, wherein the data acquisition module 10, the noise reduction module 20, the construction module 30, the optimization module 40 and the prediction module 50 are all stored in the memory as program units, and the processor executes the program units stored in the memory to realize the corresponding functions. The above modules are all located in the same processor; or, the above modules are located in different processors in any combination.

[0158] The processor includes a kernel, which retrieves the corresponding program unit from the memory. One or more kernels can be provided, and the problem that the prediction method in the prior art is difficult to effectively predict the danger of rock burst is solved by adjusting the kernel parameters.

[0159] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0160] An embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored program, wherein when the program is run, a device where the computer-readable storage medium is located is controlled to execute a method for predicting deformation of a section coal pillar.

[0161] Specifically, the deformation prediction method of the section coal pillar includes:

[0162] Step S201, obtaining the original monitoring data of the section coal pillar; step S202, determining the target monitoring data according to the original monitoring data, the empirical mode decomposition algorithm and the differential evolution algorithm; step S203, constructing the initial prediction model of the section coal pillar according to the long short-term memory neural network and the back propagation neural network, the initial prediction model having initial hyperparameters; step S204, determining the target hyperparameters according to the initial prediction model and the genetic algorithm, and obtaining the prediction training model, the genetic algorithm is used to optimize the initial hyperparameters of the initial prediction model, and the prediction training model is the initial prediction model with the target hyperparameters; step S205, determining the target prediction data of the section coal pillar according to the target monitoring data and the prediction training model.

[0163] Optionally, target monitoring data is determined according to the original monitoring data and the empirical mode decomposition algorithm and the differential evolution algorithm, including: determining the K-order modal component of the original monitoring data according to the original monitoring data and the empirical mode decomposition algorithm, the K-order modal component including multi-order components; determining the spread entropy value of the multi-order component according to the K-order modal component and the differential evolution algorithm; obtaining multiple target ratios according to the spread entropy values ​​of the multi-order components, and determining the component corresponding to the maximum ratio among the multiple target ratios as the target-order component, wherein the target ratio is the ratio of the spread entropy value of the k-th order component to the spread entropy value of the k+1-th order component, k≥1, and k is an integer; obtaining a reconstructed component, the reconstructed component being the component other than the two components corresponding to the maximum ratio in the multi-order components; and reconstructing the original monitoring data corresponding to the reconstructed component in the original monitoring data to obtain the target monitoring data.

[0164] Optionally, according to the original monitoring data and the empirical mode decomposition algorithm, the K-order modal component of the original monitoring data is determined, including: obtaining the first residual and the k+1th residual of the original monitoring data; determining the first-order component of the original monitoring data according to the first residual and a first preset formula, the first preset formula is: IMF1=x-R1, wherein IMF1 is the first-order component, x is the original monitoring data, and R1 is the first residual; determining the k+1th order component according to the first residual, the k+1th residual, and the second preset formula, the second preset formula is: IMF k+1 =R k -R k+1 , IMF k+1 is the k+1th order component, R k is the kth residual, R k+1 is the k+1th residual.

[0165] Optionally, obtaining the first residual and the k+1th residual of the original monitoring data includes: obtaining multiple original monitoring sequences, the multiple original monitoring sequences corresponding one-to-one to the multi-order components; determining the local mean of the first original monitoring sequence as the first residual, the first original monitoring sequence being obtained by adding white noise to the original monitoring data; obtaining the signal after adding white noise to the kth residual, and determining the local mean of the signal as the k+1th residual.

[0166] Optionally, according to the K-order modal components and the differential evolution algorithm, the scatter entropy values ​​of the multi-order components are determined, including: determining the target monitoring sequence according to the K-order modal components and the normal cumulative distribution function; determining multiple embedding vectors corresponding to the target monitoring sequence according to the target monitoring sequence and the embedding dimension; determining multiple probabilities of the K-order modal components according to the multiple embedding vectors and the differential evolution algorithm, and the multiple probabilities correspond one-to-one to the multi-order components; determining multiple scatter entropy values ​​corresponding one-to-one to the multi-order components according to the multiple probabilities and the scatter entropy definition.

[0167] Optionally, determining the target monitoring sequence according to the K-order modal component and the normal cumulative distribution function includes: determining a first monitoring sequence according to the K-order modal component and the normal cumulative distribution function; determining the target monitoring sequence according to the first monitoring sequence, a preset integer and a third preset formula, the third preset formula being: i =round(c*y i +0.5), where z i is the target monitoring sequence, y i is the first monitoring sequence, and c is a preset integer.

[0168] Optionally, the target monitoring data includes training set data and test set data, and the target prediction data of the section coal pillar is determined according to the target monitoring data and the prediction training model, including: determining a first prediction value according to the first historical data in the training set data and the prediction training model; judging whether the error between the first prediction value and the second historical data in the training set data is less than or equal to the target error, and obtaining a first judgment result; when the first judgment result indicates yes, stopping training the prediction training model, and determining the prediction training model as the target prediction model, and determining the target prediction data according to the test set data and the target prediction model.

[0169] An embodiment of the present invention provides an electronic device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of the method for predicting deformation of a segmented coal pillar of any one of the above embodiments of the present application are implemented.

[0170] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes any of the above-mentioned steps of the method for predicting deformation of a segmented coal pillar.

[0171] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0172] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0173] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0174] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0175] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0176] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0177] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0178] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0179] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0180] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0181] By applying the technical solution of the present application, the original monitoring data of the coal pillar of the section can be first obtained, and the initial prediction model can be constructed by the long short-term memory neural network and the back propagation neural network. Then, in order to improve the reliability of the original monitoring data, the technical solution of the present application uses the empirical mode decomposition algorithm and the differential evolution algorithm to perform noise reduction processing on the original monitoring data, and determines that the target monitoring data is obtained, and the target monitoring data is the original monitoring data after the noise reduction processing. And in order to improve the model performance of the initial prediction model, the technical solution of the present application also uses a genetic algorithm to optimize the initial hyperparameters of the initial prediction model, so that after the optimization calculation, the initial hyperparameters of the initial prediction model can be adjusted to the target hyperparameters to obtain a prediction training model. It can be understood that the prediction training model is an initial prediction model with target hyperparameters, so that the model performance of the prediction training model is better than the model performance of the initial prediction model with the initial hyperparameters. Finally, since the target monitoring data after noise reduction processing has a higher data set accuracy, the prediction training model with target hyperparameters has a higher model performance, so after the target monitoring data is input into the prediction training model, prediction data with higher prediction accuracy can be obtained, that is, the accuracy of the target prediction data of the section coal pillar is determined to be high, which has a good effect in predicting the deformation of the section coal pillar. Therefore, the technical solution of this application solves the problem that the prediction method in the prior art is difficult to effectively predict the danger of rock burst.

[0182] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting deformation of a section coal pillar, characterized in that: include: Obtaining original monitoring data of the coal pillar in the section; Determining target monitoring data according to the original monitoring data, the empirical mode decomposition algorithm and the differential evolution algorithm; According to the long short-term memory neural network and the back propagation neural network, an initial prediction model of the coal pillar in the section is constructed, wherein the initial prediction model has initial hyperparameters; According to the initial prediction model and the genetic algorithm, a target hyperparameter is determined, and a prediction training model is obtained, wherein the genetic algorithm is used to optimize the initial hyperparameter of the initial prediction model, and the prediction training model is the initial prediction model with the target hyperparameter; The target prediction data of the coal pillar in the section is determined according to the target monitoring data and the prediction training model.

2. The method according to claim 1, characterized in that Determine target monitoring data according to the original monitoring data, the empirical mode decomposition algorithm and the differential evolution algorithm, including: Determine, according to the original monitoring data and the empirical mode decomposition algorithm, a K-order modal component of the original monitoring data, wherein the K-order modal component includes multi-order components; Determining the spread entropy values ​​of the components of multiple orders according to the K-order modal components and the differential evolution algorithm; According to the spread entropy values ​​of the components of multiple orders, multiple target ratios are obtained, and the component corresponding to the maximum ratio among the multiple target ratios is determined as the target order component, wherein the target ratio is the ratio of the spread entropy value of the component of the kth order to the spread entropy value of the component of the k+1th order, k≥1, and k is an integer; Acquire a reconstructed component, where the reconstructed component is the component other than the two components corresponding to the maximum ratio among the components of the multiple orders; The original monitoring data corresponding to the reconstructed component in the original monitoring data is reconstructed to obtain the target monitoring data.

3. The method according to claim 2, characterized in that Determining, according to the original monitoring data and the empirical mode decomposition algorithm, a K-order modal component of the original monitoring data, including: Obtaining a first residual and a k+1th residual of the original monitoring data; The first-order component of the original monitoring data is determined according to the first residual and a first preset formula, wherein the first preset formula is: IMF1=x-R1, wherein IMF1 is the first-order component, x is the original monitoring data, and R1 is the first residual; Determine the k+1th order component according to the first residual, the k+1th residual and the second preset formula, wherein the second preset formula is: IMF k+1 =R k -R k+1 , IMF k+1 is the k+1th order component, R k is the kth residual, R k+1 is the k+1th residual.

4. The method according to claim 3, characterized in that Obtaining a first residual and a k+1th residual of the original monitoring data includes: Acquire a plurality of original monitoring sequences, wherein the plurality of original monitoring sequences correspond one-to-one to the components of the plurality of orders; Determine a local mean of a first original monitoring sequence as the first residual, wherein the first original monitoring sequence is obtained by adding white noise to the original monitoring data; A signal is obtained after white noise is added to the kth residual, and a local mean of the signal is determined to be the k+1th residual.

5. The method according to any one of claims 2 to 4, characterized in that Determining the distribution entropy values ​​of the multi-order components according to the K-order modal components and the differential evolution algorithm includes: Determining a target monitoring sequence according to the K-order modal component and the normal cumulative distribution function; Determining, according to the target monitoring sequence and the embedding dimension, a plurality of embedding vectors corresponding to the target monitoring sequence; Determine multiple probabilities of the K-order modal components according to the multiple embedding vectors and the differential evolution algorithm, wherein the multiple probabilities correspond one-to-one to the multiple-order components; According to the multiple probabilities and spread entropy definitions, multiple spread entropy values ​​corresponding to the multiple-order components are determined.

6. The method according to claim 5, characterized in that According to the K-order modal component and the normal cumulative distribution function, a target monitoring sequence is determined, including: Determining a first monitoring sequence according to the K-order modal component and the normal cumulative distribution function; The target monitoring sequence is determined according to the first monitoring sequence, a preset integer and a third preset formula, wherein the third preset formula is: i =round(c*y i +0.5), where z i is the target monitoring sequence, y i is the first monitoring sequence, and c is the preset integer.

7. The method according to claim 5, characterized in that The target monitoring data includes training set data and test set data. According to the target monitoring data and the prediction training model, the target prediction data of the section coal pillar is determined, including: Determining a first prediction value according to the first historical data in the training set data and the prediction training model; Determine whether an error between the first prediction value and the second historical data in the training set data is less than or equal to a target error, and obtain a first determination result; When the first judgment result indicates yes, stop training the prediction training model, determine the prediction training model as the target prediction model, and determine the target prediction data based on the test set data and the target prediction model.

8. A device for predicting deformation of a section coal pillar, characterized in that: include: A data acquisition module, used to obtain the original monitoring data of the coal pillar in the section; A noise reduction module, used to determine target monitoring data based on the original monitoring data and the empirical mode decomposition algorithm and the differential evolution algorithm; A construction module is used to construct an initial prediction model of the coal pillar in the section according to a long short-term memory neural network and a back propagation neural network, wherein the initial prediction model has initial hyperparameters; An optimization module, used to determine target hyperparameters according to the initial hyperparameters and a genetic algorithm, and obtain a prediction training model, wherein the genetic algorithm is used to perform optimization calculation on the initial hyperparameters of the initial prediction model, and the prediction training model is the initial prediction model with the target hyperparameters; The prediction module is used to determine the target prediction data of the coal pillar in the section according to the target monitoring data and the prediction training model.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the deformation prediction method for a segmented coal pillar according to any one of claims 1 to 7.

10. An electronic device, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a method for executing the deformation prediction method of the segment coal pillar described in any one of claims 1 to 7.

Images