Big data intelligent selection and design method for anti-rock burst hydraulic supports in rock burst tunnels

Abstract

The present invention provides a big data intelligent selection and design method for hydraulic supports for rock burst tunnels, which relates to the field of big data technology. This method first collects data on geological conditions, rock mechanical properties, mine working face layout, and previous hydraulic support performance and maintenance records from different mining areas and historical records. It then uses machine learning and statistical analysis methods to identify key factors affecting hydraulic support performance from the integrated data set, and then provides an intelligent hydraulic support selection system. This invention can improve the accuracy and efficiency of hydraulic support selection and ensure safety. Through in-depth analysis of large amounts of geological data, mine working face conditions, and historical hydraulic support usage, the present invention provides the most suitable hydraulic support selection recommendations for tunnels.

Description

Technical Field

[0001] The present invention relates to the field of mine big data technology, and in particular to a big data intelligent selection and design method for an anti-rock burst hydraulic support in a rock burst tunnel. Background Art

[0002] Coal mine rock burst is a dynamic phenomenon in which highly stressed coal and rock masses release energy instantaneously, often causing significant damage to the surrounding rock. Over 90% of rock bursts in my country occur in tunnels. Tunnel hydraulic support systems are the last line of defense against rock bursts, and the design of the support significantly impacts their effectiveness in resisting rock bursts.

[0003] At present, the main methods for selecting and designing hydraulic supports for mining tunnels in rock burst working faces include analytical theoretical analysis, experimental simulation, numerical calculation analysis, and empirical formulas. However, the following problems still exist:

[0004] (1) When performing numerical calculations, it is usually necessary to simplify complex geological conditions for the sake of convenience. This simplification may ignore some key factors, such as the heterogeneity of rock formations and the influence of joints and fissures, resulting in deviations between the simulation results and the actual situation;

[0005] (2) Indoor experimental test prototypes are expensive and difficult to simulate comprehensive and accurate geological information and its geomechanical processes.

[0006] (3) Empirical formulas are generally summarized from field experience, but they lack universality for projects under extreme conditions. The proposal and update cycle of empirical formulas are relatively difficult.

[0007] With the vigorous advancement of intelligent mine construction in my country, "How to achieve intelligent, safe, and efficient mining of rockburst-prone coal seams?" became the only topic in the energy and mining engineering field to be included in the China Association for Science and Technology's 2023 Top Ten Industrial Technology Issues list. Therefore, establishing a big data-driven intelligent selection and design approach is of great significance to the development of intelligent technologies and equipment for rockburst prevention and control in coal mines. Summary of the Invention

[0008] In response to the shortcomings of the existing technology, the present invention provides a big data intelligent selection and design method for anti-impact hydraulic supports in rock burst tunnels; based on an in-depth analysis of the historical usage of existing tunnel hydraulic supports, by giving corresponding geological conditions and parameter conditions in the mining process, a reasonable and safe hydraulic support selection method is given using a big data system, providing a scientific basis for the selection and design of hydraulic supports in rock burst tunnels.

[0009] A big data intelligent selection and design method for anti-rock burst hydraulic supports in rock burst tunnels includes the following steps:

[0010] Step 1: Data collection;

[0011] Step 1.1: Establish a table of geomechanical characterization parameters for rock burst working face mining tunnels;

[0012] Specifically, it includes geological factor data and mining technology factor data, among which the geological factor data include the following data: uniaxial compressive strength of coal rock σ c , coal rock impact tendency index K, coal rock elastic modulus E, internal friction angle The average ground stress value P0, mining depth h0, and the history of rock burst in the same horizontal coal seam n; mining technical factor data include the following data: the degree of pressure relief of the protective layer P1, the horizontal distance from the coal pillar left by mining the upper protective layer h1, the length of the working face L0, the width of the section coal pillar B, the thickness of the bottom coal T, the distance between the roadway excavated into the goaf, that is, the distance between the stop position and the goaf h2, and the distance between the working face advancing into the goaf, that is, the distance between the stop line and the goaf h3;

[0013] Step 1.2: Establish a table of key parameters for selecting energy-absorbing and anti-rockburst hydraulic supports in rockburst tunnels;

[0014] Specifically including the following data: initial support force F C , working resistance R w , support strength S;

[0015] Step 1.3: Collect data;

[0016] Based on the literature collection on the Internet and field investigation analysis, the geomechanical characterization parameter information of the mining roadway of the rock burst working face of group M and the key parameter information of the performance of the energy-absorbing hydraulic support in the rock burst roadway were collected and analyzed.

[0017] Step 2: Based on the neural network model, establish training samples for intelligent stent selection;

[0018] Step 2.1: Select a neural network model suitable for hydraulic support selection; the neural network model is an MLP neural network model;

[0019] Step 2.2: Define the basic structure of the neural network model;

[0020] Step 2.2.1: Determine the input layer and its size;

[0021] The geomechanical characterization parameters of the mining roadway in the rock burst working face are used as the input layer of the neural network model to calculate the following:

[0022]

[0023] Where, represents the input layer of the mth training sample of the neural network; X mRepresents the feature vector of the mth training sample;

[0024] Step 2.2.2: Determine the number of middle layers;

[0025] Assuming that the neural network model has a total of L layers, the middle layer has L-1 layers;

[0026] Step 2.2.3: Determine the number of neurons in each layer;

[0027]

[0028] In the formula is the number of neurons in the first layer; N o is the number of neurons in the output layer; N i is the number of neurons in the input layer; M is the number of samples; α is an arbitrary value variable, l≤L-1;

[0029] Step 2.2.4: Systematically construct the middle layer model;

[0030] Starting with the geomechanical characterization parameters of the rock burst working face mining tunnel, after linear transformation, the new data of the next layer is obtained by activation function processing. In this way, the key parameters for selecting the energy-absorbing and anti-bumping hydraulic support for rock burst working face are finally reflected, as shown in the following formula:

[0031]

[0032] Where W l represents the weight matrix of the lth layer of the neural network; b l Represents the bias vector of the lth layer of the neural network; Represents the output result of the mth training sample passing through the l-1th layer of the neural network; It represents the result obtained after the m-th training sample is linearly transformed in the l-1th layer of the neural network; for Middle Quantity express The result obtained after the activation function transformation;

[0033] Step 2.2.5: Set the output layer configuration of the neural network model:

[0034] The key parameters for selecting energy-absorbing and anti-bumping hydraulic supports in rock burst tunnels are used as the output layer to predict the initial support force, working resistance, and support strength of the hydraulic supports. The calculation of the output layer is expressed by the following formula:

[0035]

[0036] The ReLU activation function is used to obtain the predicted values of the key parameters for the selection of energy-absorbing and anti-impact hydraulic supports in rock burst tunnels. The specific formula is as follows:

[0037]

[0038] Where F cm , R wm , S m They represent the key parameter values for selecting the energy-absorbing and anti-rockburst hydraulic supports in the predicted rock burst tunnel, namely the initial support force, working resistance and support strength of the hydraulic supports;

[0039] Step 3: Optimize the parameter configuration of the neural network model;

[0040] Step 3.1: Calculate the loss function value;

[0041] The mean square error is selected as the loss function, and the calculation formula is:

[0042]

[0043] Z m =[F C ,R w ,S]

[0044] Where Loss is the loss function, Z m is the actual value matrix;

[0045] Step 3.2: Calculate the gradient;

[0046] Calculate the gradient of the loss function with respect to the weight matrix and bias vector;

[0047]

[0048] in, Represents the loss function Loss for the l-th layer neural network weight matrix W l The gradient, Represents the loss function Loss for the l-th layer neural network bias vector b l Gradient

[0049] Step 3.3: Iteratively optimize the neural network model parameters;

[0050] Update the weight matrix and bias vector, the calculation formula is:

[0051]

[0052] Among them, t represents the number of iterations, β represents the correction coefficient, which is used to control the update of the weight matrix W of the l-th layer neural network. l and the bias vector b of the lth layer neural network lthe step size in the process;

[0053] Repeatedly update the weight matrix and bias vector, each update t = t + 1, until the condition for stopping the iteration is:

[0054]

[0055] Where, express The infinite norm of express The infinite norm of ; ε1 and ε2 represent the set thresholds;

[0056] Step 4: Use the known geomechanical characterization parameters of the mining roadway of the rock burst working face and the mapping relationship obtained through training in steps 2 and 3 to realize intelligent selection of hydraulic supports;

[0057] Step 4.1: Real-time data collection;

[0058] With the help of the dynamic data monitoring system, the geomechanical characterization parameters of the rock burst working face mining tunnel in Table 1.1 are collected in real time and expressed as follows:

[0059] In the formula The superscript 0 in the table represents the geomechanical characterization parameters of the mining roadway corresponding to the rock burst working face collected in real time;

[0060] Step 4.2: Use the neural network model to predict the performance of the hydraulic support;

[0061] Input the data collected in step 4.1 into the neural network model trained in step 3. The trained neural network model outputs the predicted key parameters for selecting the energy-absorbing and anti-impact hydraulic support for rock burst tunnels. Output value Used to guide the intelligent selection of hydraulic supports.

[0062] The beneficial effects of adopting the above technical solution are:

[0063] The present invention provides a big data-based intelligent selection and design method for hydraulic supports to prevent rock bursts in roadways. This method provides recommendations for selecting the most suitable hydraulic supports for roadways through in-depth analysis of extensive geological data, mine face conditions, and historical hydraulic support usage. Data on geological conditions, rock mechanical properties, mine face layout, and historical hydraulic support performance and maintenance records are collected from different mining areas and historical records. Machine learning and statistical analysis methods are then used to identify key factors influencing hydraulic support performance from this integrated data set. This system then provides an intelligent hydraulic support selection system. This invention improves the accuracy and efficiency of hydraulic support selection, ensuring safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is an overall flow chart of the intelligent selection and design method for anti-rock pressure hydraulic supports in rock burst tunnels according to an embodiment of the present invention;

[0065] Figure 2 This is a neural network model diagram in an embodiment of the present invention. DETAILED DESCRIPTION

[0066] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention but are not intended to limit the scope of the present invention.

[0067] This example is aimed at the 1206 working face of a certain mine, and its coal rock uniaxial compressive strength σ c =3.09MPa, coal rock impact tendency index K = 2.196, coal rock elastic modulus E = 15.7GPa, internal friction angle The average ground stress value P0 is 27.01 MPa, the mining depth h0 is 752.5 m, the number of rock bursts in the same horizontal coal seam is n=0, and the pressure relief degree of the protective layer is average (the pressure relief degree of the protective layer is divided into "good, medium, average, and very poor", and the P1 values in these four cases are set to 0, 1, 2, and 3 respectively). Then, in this embodiment, P1 is 2, the horizontal distance from the coal pillar left by mining the upper protective layer is h1=60 m, the working face length L0=150 m, the section coal pillar width B=25 m, the bottom coal thickness T=800 mm, the distance between the stop position of the tunnel excavated into the goaf and the goaf is h2=50 m, and the distance between the stop line and the goaf of the working face advancing into the goaf is h3=40 m.

[0068] A big data intelligent selection and design method for anti-rock pressure hydraulic supports in rock burst tunnels, such as Figure 1 As shown, the following steps are included:

[0069] Step 1: Data collection;

[0070] Step 1.1: Establish a table of geomechanical characterization parameters for rock burst working face mining tunnels;

[0071]

[0072]

[0073] Step 1.2: Establish a table of key parameters for selecting energy-absorbing and anti-rockburst hydraulic supports in rockburst tunnels;

[0074] Serial number Parameter name symbol Specific meaning of parameters 1 Initial support force <![CDATA[F C (10 6 N)]]> The force exerted by the hydraulic support when it first contacts the roof 2 Working resistance <![CDATA[R w (10 6 N)]]> The maximum supporting force of the hydraulic support when it withstands the pressure from the top plate 3 Support strength S(MPa) The support capacity that hydraulic supports can provide per unit area

[0075] Step 1.3: Collect data;

[0076] Based on the literature collection on the Internet and field investigation analysis, the geomechanical characterization parameter information of the mining roadway of the rock burst working face of group M and the key parameter information of the performance of the energy-absorbing hydraulic support in the rock burst roadway were collected and analyzed.

[0077] In this example, 200 sets of data σ for establishing geomechanical characterization parameters of rock burst working face mining tunnels were collected through field investigation and literature search. c ,K,E, P0, h0, n, P1, h1, L0, B, T, h2, h3 and the corresponding key parameter data for selecting energy-absorbing and anti-impact hydraulic supports for rock burst tunnels F C ,R w ,S and construct the sample matrix. The samples are divided into three parts: training set, validation set and test set. The number of samples is 160 groups, 40 groups and 40 groups respectively. The training set and validation set are normalized.

[0078] Step 2: Based on the neural network model, such as Figure 2 As shown, a training sample for intelligent selection of stents is established;

[0079] Step 2.1: Select a neural network model suitable for hydraulic support selection; the neural network model is an MLP neural network model;

[0080] Step 2.2: Define the basic structure of the neural network model;

[0081] Step 2.2.1: Determine the input layer and its size;

[0082] The geomechanical characterization parameters of the mining roadway in the rock burst working face are used as the input layer of the neural network model to calculate the following:

[0083]

[0084] Where, represents the input layer of the mth training sample of the neural network; X m Represents the feature vector of the mth training sample;

[0085] Step 2.2.2: Determine the number of middle layers;

[0086] Assuming that the neural network model has a total of L layers, the middle layer has L-1 layers. In this embodiment, based on factors such as the number of data types, the neural network is assumed to have a total of 4 layers, and the middle layer has 3 layers.

[0087] Step 2.2.3: Determine the number of neurons in each layer;

[0088] According to the "empirical formula" proposed by stackoverflow;

[0089]

[0090] In the formula is the number of neurons in the first layer; N o is the number of neurons in the output layer; N i is the number of neurons in the input layer; M is the number of samples; α is a variable that can take any value, usually 2-10, l≤L-1; in this embodiment, the value of α is 4, and the number of neurons in each layer is

[0091] Step 2.2.4: Systematically construct the middle layer model;

[0092] In a neural network, the core of the forward propagation process of information lies in the continuous transformation between layers. According to this technical solution, the geomechanical characterization parameters of the rock burst working face mining roadway are used as the starting point. After linear transformation, the activation function is used to process the new data of the next layer. In this way, the data is passed layer by layer, and finally the key parameters for the selection of energy-absorbing and anti-impact hydraulic supports for rock burst working face are reflected, as shown in the following formula:

[0093]

[0094] Where W l represents the weight matrix of the lth layer of the neural network; b l Represents the bias vector of the lth layer of the neural network; Represents the output result of the mth training sample passing through the l-1th layer of the neural network; It represents the result obtained after the m-th training sample is linearly transformed in the l-1th layer of the neural network; for Middle Quantity express The result obtained after the activation function transformation;

[0095] Step 2.2.5: Set the output layer configuration of the neural network model:

[0096] The key parameters for selecting energy-absorbing and anti-bumping hydraulic supports in rock burst tunnels are used as the output layer to predict the initial support force, working resistance, and support strength of the hydraulic supports. The calculation of the output layer is expressed by the following formula:

[0097]

[0098] The ReLU activation function is used to obtain the predicted values of the key parameters for the selection of energy-absorbing and anti-impact hydraulic supports in rock burst tunnels. The specific formula is as follows:

[0099]

[0100] Where F cm , R wm , S m They represent the key parameter values for selecting the energy-absorbing and anti-rockburst hydraulic supports in the predicted rock burst tunnel, namely the initial support force, working resistance and support strength of the hydraulic supports;

[0101] Step 3: Optimize the parameter configuration of the neural network model;

[0102] Step 3.1: Calculate the loss function value;

[0103] The mean square error is selected as the loss function, and the calculation formula is:

[0104]

[0105] Z m =[F C ,R w ,S]

[0106] Where Loss is the loss function, Z m is the actual value matrix;

[0107] Step 3.2: Calculate the gradient;

[0108] Calculate the gradient of the loss function with respect to the weight matrix and bias vector;

[0109]

[0110]

[0111] in, Represents the loss function Loss for the l-th layer neural network weight matrix W l The gradient, Represents the loss function Loss for the l-th layer neural network bias vector b l Gradient

[0112] Step 3.3: Iteratively optimize the neural network model parameters;

[0113] Update the weight matrix and bias vector, the calculation formula is:

[0114]

[0115] Among them, t represents the number of iterations, β represents the correction coefficient, which is used to control the update of the weight matrix W of the l-th layer neural network. l and the bias vector b of the lth layer neural network l The step size in the process; in this embodiment, the number of iterations t=100;

[0116] Repeatedly update the weight matrix and bias vector, each update t = t + 1, until the condition for stopping the iteration is:

[0117]

[0118] Where, express The infinite norm of express The infinite norm of ; ε1 and ε2 represent the set thresholds;

[0119] Step 4: Use the known geomechanical characterization parameters of the mining roadway of the rock burst working face and the mapping relationship obtained through training in steps 2 and 3 to realize intelligent selection of hydraulic supports;

[0120] Step 4.1: Real-time data collection;

[0121] With the help of the dynamic data monitoring system, the geomechanical characterization parameters of the rock burst working face mining tunnel in Table 1.1 are collected in real time and expressed as follows:

[0122] In the formula The superscript 0 in the table represents the geomechanical characterization parameters of the mining roadway corresponding to the rock burst working face collected in real time;

[0123] In this embodiment, the uniaxial compressive strength of coal rock, coal rock impact tendency index, coal rock elastic modulus, and internal friction angle are obtained through indoor testing using a testing machine; the degree of pressure relief of the protective layer is obtained using an existing stress gauge; the mean ground stress is obtained using the stress contact method ground stress testing method, the mining depth is recorded on the drawings, and the historical number of rock bursts in the same horizontal coal seam, that is, the record information of rock bursts; the degree of pressure relief of the protective layer, the horizontal distance from the coal pillar left by mining the upper protective layer, the length of the working face, the width of the section coal pillar, the thickness of the bottom coal, the roadway excavated into the goaf, the distance between the stop excavation position and the goaf, the working face advancing into the goaf, and the distance between the stop mining line and the goaf can be obtained by measuring the mining project plan;

[0124] Step 4.2: Use the neural network model to predict the performance of the hydraulic support;

[0125] Input the data collected in step 4.1 into the neural network model trained in step 3. The trained neural network model outputs the predicted key parameters for selecting the energy-absorbing and anti-impact hydraulic support for rock burst tunnels. Output value Used to guide the intelligent selection of hydraulic supports.

[0126] Using the data from the example, write the data in the dataset in a table format and give the specific values involved in the above process;

[0127]

[0128] The specific values involved in the above process are;

[0129] Number of neurons in each layer Number of layers L = 4; number of iterations t = 100; threshold ε1 = ε2 = 1*10 -3

[0130] The weights of each layer:

[0131]

[0132] Bias vectors for each layer:

[0133] b 1 =[-0.1649 -0.2819 -0.2472]

[0134] b 2 =[-0.2326 0.0803 0.3128]

[0135] b 3 =[0.2840 0.3948 -0.4565]

[0136] b 4 =[0.7036 0.6204 0.2743]

[0137] Real-time data with sequence number 1 K 1 =2.196, E 1 =15.7, n 1 =0, B 1 =25, T 1 =

[0138] 800, For example, the prediction result is:

[0139] Real-time data with sequence number 2 K 2 =2.25, E 2 =16.0, n 2 =1, B 2 =30, T 2 =850, For example, the prediction result is:

[0140] Real-time data with sequence number 3 K 3 =2.05, E 3 =15.5, n 3 =0, B 3 =20, T 3 =780, For example, the prediction result is:

[0141] The key parameters for selecting hydraulic supports for energy absorption and anti-impact rock burst tunnels are predicted to be:

[0142] No. 1 Real-time data prediction: initial support force Working resistance Support strength

[0143] No. 2 Real-time Data Forecast: Initial Support Force Working resistance Support strength

[0144] No. 3 Real-time data prediction: initial support force Working resistance Support strength

[0145] The above description is merely a preferred embodiment of the present disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also encompass other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the above-mentioned inventive concept. For example, a technical solution formed by mutually replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in the embodiments of the present disclosure.

Claims

1. A big data intelligent selection and design method for anti-rock pressure hydraulic supports in rock burst tunnels, characterized in that: The following steps are involved: Step 1: Data collection; Step 2: Based on the neural network model, establish training samples for intelligent stent selection; Step 2.1: Select a neural network model suitable for hydraulic support selection; the neural network model is an MLP neural network model; Step 2.2: Define the basic structure of the neural network model; Step 2.2.1: Determine the input layer and its size; The geomechanical characterization parameters of the mining roadway in the rock burst working face are used as the input layer of the neural network model to calculate the following: Where, represents the input layer of the mth training sample of the neural network; X m Represents the feature vector of the mth training sample; Step 2.2.2: Determine the number of middle layers; Assuming that the neural network model has a total of L layers, the middle layer has L-1 layers; Step 2.2.3: Determine the number of neurons in each layer; In the formula is the number of neurons in the first layer; N o is the number of neurons in the output layer; N i is the number of neurons in the input layer; M is the number of samples; α is an arbitrary value variable, l≤L-1; Step 2.2.4: Systematically construct the middle layer model; Starting with the geomechanical characterization parameters of the rock burst working face mining tunnel, after linear transformation, the new data of the next layer is obtained by activation function processing. In this way, the key parameters for selecting the energy-absorbing and anti-bumping hydraulic support for rock burst working face are finally reflected, as shown in the following formula: Where W l represents the weight matrix of the lth layer of the neural network; b l Represents the bias vector of the lth layer of the neural network; Represents the output result of the mth training sample passing through the l-1th layer of the neural network; It represents the result obtained after the m-th training sample is linearly transformed in the l-1th layer of the neural network; for Middle Quantity express The result obtained after the activation function transformation; Step 2.2.5: Set the output layer configuration of the neural network model: The key parameters for selecting energy-absorbing and anti-bumping hydraulic supports in rock burst tunnels are used as the output layer to predict the initial support force, working resistance, and support strength of the hydraulic supports. The calculation of the output layer is expressed by the following formula: The ReLU activation function is used to obtain the predicted values of the key parameters for the selection of energy-absorbing and anti-impact hydraulic supports in rock burst tunnels. The specific formula is as follows: Where F cm , R wm , S m They represent the key parameter values for selecting the energy-absorbing and anti-rockburst hydraulic supports in the predicted rock burst tunnel, namely the initial support force, working resistance and support strength of the hydraulic supports; Step 3: Optimize the parameter configuration of the neural network model; Step 4: The known geomechanical characterization parameters of the rock burst working face mining tunnel are used in the mapping relationship trained from Step 2 to Step 3 to realize the intelligent selection of hydraulic supports.

2. The big data intelligent selection and design method for rock burst tunnel anti-rock burst hydraulic support according to claim 1 is characterized in that: The step 1 specifically includes the following steps: Step 1.1: Establish a table of geomechanical characterization parameters for rock burst working face mining tunnels; Specifically, it includes geological factor data and mining technology factor data, among which the geological factor data include the following data: uniaxial compressive strength of coal rock σ c , coal rock impact tendency index K, coal rock elastic modulus E, internal friction angle The average ground stress value P0, mining depth h0, and the history of rock burst in the same horizontal coal seam n; mining technical factor data include the following data: the degree of pressure relief of the protective layer P1, the horizontal distance from the coal pillar left by mining the upper protective layer h1, the length of the working face L0, the width of the section coal pillar B, the thickness of the bottom coal T, the distance between the roadway excavated into the goaf (i.e., the distance between the stop position and the goaf H2), and the distance between the working face advancing into the goaf (i.e., the distance between the stop line and the goaf H3); Step 1.2: Establish a table of key parameters for selecting energy-absorbing and anti-rockburst hydraulic supports in rockburst tunnels; Specifically including the following data: initial support force F C , working resistance R w , support strength S; Step 1.3: Collect data; Based on the literature collection on the Internet and field investigation analysis, the geomechanical characterization parameter information of the mining roadway of the rock burst working face of group M and the key parameter information of the performance of the energy-absorbing hydraulic support in the rock burst roadway were collected and analyzed.

3. The big data intelligent selection and design method for rock burst tunnel anti-rock burst hydraulic support according to claim 1 is characterized in that: The step 3 specifically includes the following steps: Step 3.1: Calculate the loss function value; The mean square error is selected as the loss function, and the calculation formula is: Z m =[F C ,R w ,S] Where Loss is the loss function, Z m is the actual value matrix; Step 3.2: Calculate the gradient; Calculate the gradient of the loss function with respect to the weight matrix and bias vector; in, Represents the loss function Loss for the l-th layer neural network weight matrix W l The gradient, Represents the loss function Loss for the l-th layer neural network bias vector b l Gradient Step 3.3: Iteratively optimize the neural network model parameters; Update the weight matrix and bias vector, the calculation formula is: Among them, t represents the number of iterations, β represents the correction coefficient, which is used to control the update of the weight matrix W of the l-th layer neural network. l and the bias vector b of the lth layer neural network l the step size in the process; Repeatedly update the weight matrix and bias vector, each update t = t + 1, until the condition for stopping the iteration is: Where, express The infinite norm of express The infinite norm of ; ε1 and ε2 represent the set thresholds.

4. The big data intelligent selection and design method for rock burst tunnel anti-rock burst hydraulic support according to claim 1 is characterized in that: The step 4 specifically includes the following steps: Step 4.1: Real-time data collection; With the help of the dynamic data monitoring system, the geomechanical characterization parameters of the rock burst working face mining tunnel in Table 1.1 are collected in real time and expressed as follows: In the formula The superscript 0 in the table represents the geomechanical characterization parameters of the mining roadway corresponding to the rock burst working face collected in real time; Step 4.2: Use the neural network model to predict the performance of the hydraulic support; Input the data collected in step 4.1 into the neural network model trained in step 3. The trained neural network model outputs the predicted key parameters for selecting the energy-absorbing and anti-impact hydraulic support for rock burst tunnels. Output value Used to guide the intelligent selection of hydraulic supports.

Images