A coal mine underground drilling process working condition recognition method based on gradient boosting decision

Abstract

The application discloses a kind of based on gradient lifting decision coal mine underground drilling process working condition identification method, comprising the following steps: step 1: coal mine drilling process operating state grade division: the state grade division corresponding to process capability index is established, i.e. normal, conservative, three kinds of working conditions of overexcited;Step 2: coal mine drilling process working condition identification model parameter selection: select fixed threshold drilling operating parameter as the input parameter of drilling process working condition identification model;Step 3: coal mine drilling process working condition identification model realization: gradient lifting decision tree method is used to realize model, finally realizes coal mine drilling process working condition identification.The method of the application overcomes the limitations brought by regional restrictions in traditional methods, not only improves the accuracy of coal mine drilling process working condition identification result, but also provides guidance for the research of coal mine drilling process state monitoring, performance optimization and intelligent control, with practicality and applicability.

Description

Technical Field

[0001] This invention relates to the field of coal mine drilling engineering, and more particularly to a method for identifying working conditions during underground coal mine drilling based on gradient boosting decision-making. Background Technology

[0002] Coal is my country's primary energy source and important raw material, providing a large amount of primary energy for national economic and social development and supporting the average annual growth of GDP. However, the underground working environment in coal mines is harsh and unsafe. Furthermore, due to limitations in current technology, there is a significant lack of information regarding the drilling process status. Drillers lack real-time awareness of the drilling conditions, leading to technical problems such as low drilling efficiency, decreased performance of downhole drill string assemblies, frequent in-hole accidents, and difficulty in ensuring the pre-set wellbore trajectory. Drilling process condition identification technology can accurately reflect the operating status of the drilling system, providing reliable operational guidance for drillers. When drillers can identify problems in the current operating status, they can adjust operating parameters in a timely manner, ensuring the safe, efficient, and stable progress of the drilling process.

[0003] In recent years, with the accelerated development of industrial digitalization and intelligentization, numerous data-driven methods have been applied to dynamic process monitoring and process modeling in the coal industry. Many scholars and drilling companies both domestically and internationally have dedicated themselves to using various methods and technologies to monitor and analyze the trends in coal mine drilling processes, analyze and extract influencing factors, identify the operational status of coal mine drilling, and provide suggestions for saving drilling costs and improving drilling efficiency. Therefore, this invention introduces a process capability index as a benchmark for measuring the operational status of the drilling process, thereby enabling the identification of working conditions in complex coal mine drilling processes. The proposed method provides insights into the identification of working conditions in coal mine tunnel drilling processes, guiding drillers to adjust parameters in a timely manner and maintain safe and efficient drilling. Summary of the Invention

[0004] This invention proposes a method for identifying the operating conditions of underground coal mine drilling based on gradient boosting decision-making. It introduces a process capability index as an indicator to measure the operating status of coal mine drilling, overcoming the limitations of traditional methods due to geological factors. This method aims to achieve better identification of the operating conditions of coal mine drilling and ensure the safe, efficient, and stable operation of the drilling process.

[0005] To achieve the above objectives, the technical solution adopted by the present invention includes:

[0006] A method for identifying working conditions during underground drilling in coal mines based on gradient boosting decision-making includes the following steps:

[0007] Step 1: Classification of Operating Status Levels in Coal Mine Drilling Process: Introducing the Process Capability Index C pUsing drilling speed V as the calculation variable, the operating status level G of the coal mine drilling process is obtained, and the status level classification corresponding to the process capability index is established, namely, normal, conservative and excessive working conditions.

[0008] Step 2: Parameter selection for coal mine drilling process condition identification model: After filtering and normalizing the drilling operation parameters, the singular value decomposition method is used to extract variable features of the drilling operation parameters. The inflection point of the singular value fitting curve is used as the basis for threshold setting. Drilling operation parameters with fixed thresholds are selected as input parameters for the drilling process condition identification model.

[0009] Step 3: Implementation of the coal mine drilling process condition identification model: Using the singular vector of the input parameters obtained in Step 2 as the model input and the operating status level of the coal mine drilling process as the model output, the gradient boosting decision tree method is used to implement the model, and finally realize the identification of the coal mine drilling process conditions.

[0010] Optionally, step 1 specifically includes:

[0011] 1.1 Based on the operating status of the coal mine drilling process, analyze the changes in drilling process operating parameters and obtain the distribution of drilling speed, i.e. The process capability index value is obtained by using the drilling rate as the calculation variable, along with the target drilling rate value T. The Taguchi process capability index calculation formula used is as follows:

[0012]

[0013] Where USL and LSL are the upper and lower tolerance limits of the selected sliding window, respectively, and the corrected distribution variance.

[0014] 1.2 Process capability index and corresponding performance level are as follows: When C p ≥1.67 indicates an overexcitation state; when 1≤C p <1.67 is considered normal; when C p <1 indicates a conservative state.

[0015] Optionally, step 2 specifically includes:

[0016] 2.1 The maximum and minimum value normalization unified parameter format is adopted to filter and normalize the coal mine drilling process data to obtain standardized coal mine drilling data, and the singular value decomposition (SVD) method is used to extract variable features.

[0017] I = USV T ;

[0018] If I is an m*n matrix, then U is an m*n unitary matrix; S is an m*n diagonal matrix whose diagonal numbers are real and non-negative; V is an n*n unitary matrix, where VT Denote the conjugate transpose of V;

[0019] The singular value λ is from II T and I T The non - zero square roots of the eigenvalues obtained from I give two singular vectors; the columns of U and V are called left singular vectors and right singular vectors respectively; the arrangement of the singular values in the diagonal matrix S is:

[0020]

[0021] where λ1≥λ2≥...≥λ r ,λ r >0 (i = 1,2,...,r) are the singular values of matrix I;

[0022] Screen the singular values according to a fixed threshold; rearrange the non - zero singular values and select a fixed number i of singular values from large to small, 0 < i < r; the inflection point of the singular value fitting curve is used as the basis for threshold setting, and the coal mine drilling parameters with i = 3 are selected as the input parameters of the model;

[0023] 2.2 According to the above judgment criteria, select the drilling characteristic parameters of the three characteristics of feed pressure, torque and power head displacement in the coal mine drilling parameters as the input parameters of the model.

[0024] Optionally, step 3 specifically includes:

[0025] 3.1 Use the singular vectors of feed pressure, torque and power head displacement as the input of the model; divide according to the state level corresponding to the process capability index, and add state labels to the operation data as the output of the model;

[0026] 3.2 Use the GBDT classification algorithm to construct the model, and its implementation method is as follows:

[0027] Define the training set as D = {(x1,y1),(x2,y2),…,(x m ,y m )}, where m is the number of samples, the finally obtained strong learner is F(x), and the loss function is;

[0028]

[0029] where y k = {0,1} indicates whether it belongs to the k - th class, 0 means no, 1 means yes; k = 1,2,…,K, K; is the number of operation state level categories 3; P k (x) represents the probability that the sample point belongs to each class.

[0030] Optionally, the specific steps for constructing the model using the GBDT classification algorithm are:

[0031] Step 3.1: Initialize the weak learner;

[0032] F k0 (x) = 0, k = 1, 2, ..., K;

[0033] Step 3.2: At the t-th iteration:

[0034] Step 3.2-1: Calculate the probability that a sample point belongs to each category;

[0035]

[0036] Step 3.2-2: For each category k = 1, 2, ..., K;

[0037] Step 3.2-2-1: Calculate the negative gradient error;

[0038] r ki =y ki -p k (x i ), i = 1, 2, ..., m;

[0039] Step 3.2-2-2: Using {(x1,rt1),…,(xm,rtm)} as training data, fit the t-th regression tree, Rt tlj For each leaf node region, j = 1, 2, ..., J, where J is the number of leaf nodes, calculate the best fit value for each leaf node region.

[0040]

[0041] Step 3.2-2-3: Update the strong learner;

[0042]

[0043] Step 3.3: The iteration is complete, and the final strong learner is obtained;

[0044]

[0045] The final result F Tk (x) can be used to calculate the probability of being classified into the k-th class;

[0046]

[0047]

[0048] The final output class is c(k,k′), which is the joint cost of predicting the class as the kth class when the true value is k′. That is, the class with the highest probability is the output class.

[0049] The technical effects of this invention are as follows:

[0050] Introducing a process capability index as an indicator to measure the operational status of coal mine drilling overcomes the limitations of traditional methods imposed by geological factors, offering better universality and meeting actual field needs. Simultaneously, standardized data is obtained by filtering and normalizing coal mine drilling data. Then, the SVD method is used to extract features from this standardized data, and singular vectors are selected based on a fixed threshold as input parameters for the drilling process condition identification model. Finally, because Gradient Boosting Decision Technology (GBDT) can flexibly handle various data types, requires less feature engineering, and has high prediction accuracy and robustness, the GBDT method is used to implement the drilling process condition identification model. Verification with actual data shows that the invented condition identification method performs well in identifying the operational status of coal mine drilling processes. Attached Figure Description

[0051] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings:

[0052] Figure 1 It is a curve plot of the singular values ​​of the variables;

[0053] Figure 2 These are the implementation steps of the coal mine drilling process condition identification method of the present invention;

[0054] Figure 3 This is a flowchart of the coal mine drilling process condition identification method designed in this invention;

[0055] Figure 4 This is the result of the coal mine drilling process condition identification in this invention. Detailed Implementation

[0056] To provide a clearer understanding of the technical features, objectives, and effects of the present invention, specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0057] This invention addresses actual field needs by first introducing a process capability index as a benchmark for measuring the operating status of the drilling process, defining different levels for the actual operating status of the drilling process; secondly, filtering and normalizing coal mine drilling data to obtain standardized data; then, using the SVD method to extract features from the standardized coal mine drilling data, and selecting singular vectors based on a fixed threshold as input parameters for the drilling process condition identification model; finally, establishing the drilling process condition identification model using the GBDT method, with the model outputting the current operating status (i.e., the current working condition), thus completing the identification of the coal mine drilling process conditions.

[0058] The method for identifying working conditions in underground coal mine drilling based on gradient boosting decision-making of the present invention mainly includes the following steps:

[0059] Step 1: Classification of Operating Status Levels in Coal Mine Drilling Process: Introducing the Process Capability Index C p Using drilling speed V as the calculation variable, the operating status level G of the coal mine drilling process is obtained, and the status level classification corresponding to the process capability index is established, namely, normal, conservative and excessive working conditions.

[0060] Step 2: Parameter selection for coal mine drilling process condition identification model: After filtering and normalizing the drilling operation parameters, the Singular Value Decomposition (SVD) method is used to extract features from the standardized coal mine drilling data. The inflection point of the singular value fitting curve is used as the basis for threshold setting, and a singular vector with a fixed threshold is selected as the input parameter for the drilling process condition identification model.

[0061] Step 3: Implementation of the coal mine drilling process condition identification model: The model takes a selected singular vector with a fixed threshold as the model input and the coal mine drilling process operation status level (i.e., drilling condition) defined according to the process capability index as the model output. The gradient boosting decision tree (GBDT) method is used to implement the model, and finally realizes the identification of coal mine drilling process conditions.

[0062] refer to Figure 2 , Figure 2 This is a flowchart of the coal mine drilling process condition identification method of the present invention. The present invention specifically includes the following steps:

[0063] Step 1: Classification of Operating Status Levels in Coal Mine Drilling Process

[0064] The operational capacity of the coal mine drilling process reflects the drilling efficiency. This is illustrated by introducing the process capacity index C. p Using drilling speed V as the calculation variable, the operating level G of the coal mine drilling process is obtained, and the performance level classification corresponding to the process capability index is established.

[0065] (1) The process capability index refers to the degree to which process capability meets product quality standard requirements (specification range, etc.), also known as the process capability index. It refers to the actual processing capability of a process under controlled (stable) conditions over a certain period of time. It is the inherent capability of the process, or in other words, the process's ability to guarantee quality. The process referred to here is the process involving the combined effects of five basic quality factors: operator, machine, raw materials, process method, and production environment, which is the production process of product quality. The coal mine drilling process falls within the application scope of the process capability index, therefore, the process capability index is selected for classification. Based on the operating status of the coal mine drilling process, the changes in operating parameters are analyzed and the distribution of drilling speed is obtained, i.e.; The target drilling rate value T is used; a sliding window of length 10 is employed, with drilling rate as the calculation variable, to obtain the corresponding process capability index value; the Taguchi process capability index calculation formula used in this invention is as follows:

[0066]

[0067] Where USL and LSL are the upper and lower tolerance limits of the selected sliding window, respectively; the corrected distribution variance It conforms to a Gaussian distribution. μ represents the mean. Indicates variance.

[0068] (2) Based on the calculated process capability index value, the capability level classification is shown in Table 1.

[0069] Table 1 Evaluation Criteria for Process Capability Index

[0070] <![CDATA[C p Range of values level Coal mine drilling conditions <![CDATA[C p ≥1.67]]> <![CDATA[G1]]> overexcitement <![CDATA[1≤ p <1.67]]> <![CDATA[G2]]> Normal state <![CDATA[C p <1]]> <![CDATA[G3]]> conservative state

[0071] Step 2: Parameter selection for the coal mine drilling process condition identification model

[0072] The coal mine drilling process data is filtered and normalized using a maximum-minimum value normalization format. Singular Value Decomposition (SVD) is used to extract features from the standardized coal mine drilling data. The inflection point of the singular value fitting curve is used as the basis for threshold setting, with a fixed threshold value of 3. The first three singular variables are selected as feature variables as input parameters for the working condition identification model. Step 2, the parameter selection for the working condition identification model, includes the following steps:

[0073] (1) Based on the obtained standardized coal mine drilling data, the SVD method was used to extract variable features, and the first three singular vectors were selected as input parameters for the working condition identification model.

[0074] Singular value decomposition (SVD) is the factorization of any real or complex matrix. It is a linear algebraic method used to decompose a matrix I into its constituent matrices U, S, and V. Mathematically:

[0075] I = USV T ;

[0076] If I is an m*n matrix, then U is an m*n unitary matrix. S is an m*n diagonal matrix whose diagonal numbers are real and non-negative. V is an n*n unitary matrix, where Vn... T This represents the conjugate transpose of V. Using this decomposition expressed in the equation above, the process is called singular value decomposition. The elements of matrix S are arranged in descending order of size, decreasing along the diagonal, with the highest value in the upper left part of the matrix and the lowest value in the lower right part; these elements are called singular values. The singular value λ is essentially derived from II. T and I TThe non-zero square roots of the eigenvalues ​​obtained by I are thus derived. Two singular vectors, U and V, are obtained; their columns are called the left and right singular vectors, respectively. The arrangement of the singular values ​​in the diagonal matrix S is as follows:

[0077]

[0078] Where λ1≥λ2≥...≥λ r ,λ r >0 (i = 1, 2, ..., r) are the singular values ​​of matrix I.

[0079] The inflection points of the singular value fitting curves are used as the basis for threshold setting, and singular values ​​are filtered according to a fixed threshold. Non-zero singular values ​​are rearranged, and a fixed number i of singular values ​​are selected from largest to smallest, 0... <i<r。

[0080] (2) Based on the above judgment criteria, the singular variables of three characteristics of coal mine drilling parameters, namely feed pressure, torque and power head displacement, are selected as input parameters of the working condition identification model.

[0081] Step 3: Implementation of the Coal Mine Drilling Process Condition Identification Model

[0082] Using the singular variables selected in step 2 as model inputs and the coal mine drilling process operating status levels defined by the process capability index in step 1 as model outputs, the GBDT method is employed to implement the model, thereby achieving coal mine drilling process operating condition identification. The implementation of the coal mine drilling process operating condition identification model includes the following sub-steps:

[0083] (1) Use the singular variables selected in step 2 as model input; use the operation status level classification method to add status labels to the operation data as the output of the model, that is, output the operation status level (current working condition) of the coal mine drilling process.

[0084] (2) Gradient Boosting Decision Tree (GBDT) can flexibly handle various types of data, requires less feature engineering, and has high prediction accuracy and good robustness. GBDT is an ensemble learning decision tree algorithm that uses the boosting idea to combine multiple weak learners (CART regression trees) into a strong learner. Using the gradient descent method, through continuous iterative learning, the output residual of the strong learner in the previous cycle is used to train the weak learner to fit the current cycle, thereby continuously reducing the residual and improving the model accuracy. The GBDT algorithm includes two types: regression algorithm and classification algorithm. The main difference lies in their loss functions. The regression algorithm generally uses the squared difference function, while the classification algorithm uses a log-likelihood function similar to logistic regression. The coal mine drilling process operation status perception model outputs operation status level labels, which is essentially a multi-classifier. Therefore, the GBDT classification algorithm is used for model construction, and its implementation method is as follows:

[0085] Define the training set as D = {(x1,y1),(x2,y2),…,(x...} m ,y m )}, where m is the number of samples, and the final strong learner is F(x), with the loss function being:

[0086]

[0087] Where y k ={0,1} indicates whether it belongs to the k-th class, where 0 means no and 1 means yes. k = 1,2,…,K, where K is the number of running status level categories in this paper, which is 3.

[0088] The specific steps of the algorithm are as follows:

[0089] Step 3.1: Initialize the weak learner;

[0090] F k0 (x) = 0, k = 1, 2, ..., K;

[0091] Step 3.2: At the t-th iteration:

[0092] Step 3.2-1: Calculate the probability that a sample point belongs to each category.

[0093]

[0094] Step 3.2-2: For each category k = 1, 2, ..., K;

[0095] Step 3.2-2-1: Calculate the negative gradient error:

[0096] r ki =y ki -p k (x i ), i = 1, 2, ..., m;

[0097] Step 3.2-2-2: 2: Set {(x1,r t1 ),…,(x m ,r tm Using )} as training data, the t-th regression tree is fitted, R tkj For each leaf node region, j = 1, 2, ..., J, where J is the number of leaf nodes, calculate the best fit value for each leaf node region.

[0098]

[0099] Step 3.2-2-3: Update the strong learner;

[0100]

[0101] Step 3.3: The iteration is complete, and the final strong learner is obtained;

[0102]

[0103] The final result F k (x) can be used to calculate the probability of being classified into the k-th class;

[0104]

[0105] Since the classifier uses the difference between the predicted probability values ​​and the true probability values ​​of k categories to fit the loss, the probabilities must finally be converted into the actual output categories:

[0106]

[0107] The final output class is c(k,k′), which is the joint cost of predicting the class as the kth class when the true value is k′. That is, the class with the highest probability is the output class.

[0108] The working condition identification model established using the above method takes singular variables representing the operating parameters of the coal mine drilling process as input and outputs the operating status level, thereby achieving the purpose of identifying the working conditions of the coal mine drilling process. The drilling process working condition identification process is as follows: Figure 3 As shown.

[0109] This implementation uses 112 sets of coal mine drilling data from a specific well site as the specific object, including key operating variables such as power head displacement, feed pressure, power head speed, torque, and drilling speed. Due to the lack of measured torque data, and considering the positive correlation between torque and main pump pressure, main pump pressure is used as a substitute for torque. Based on this, a process capability index is first introduced to classify the coal mine drilling operation status, establishing a corresponding level classification. Secondly, the coal mine drilling data is filtered and normalized to obtain standardized coal mine drilling data. Then, the SVD method is used to extract features from the standardized coal mine drilling data, and the selected singular variables are used as input parameters for the working condition identification model. Finally, the GBDT method is used to establish the working condition identification model, which outputs the current operating status (i.e., the current working condition), completing the working condition identification of the coal mine drilling process. The specific steps are as follows:

[0110] (1) Classification of operational status levels in coal mine drilling process

[0111] During coal mine drilling, drilling efficiency continuously decreases as the encountered strata change, failing to meet pre-set drilling requirements and resulting in wasted costs. Therefore, determining the current operating status of coal mine drilling based on drilling process state parameters is crucial for optimizing and controlling the drilling process. Traditional drilling rate equation methods cannot overcome evaluation errors caused by geographical limitations. This invention introduces a process capability index to establish a method for identifying the operating conditions of coal mine underground drilling processes based on gradient boosting decision-making, aiming to achieve the goal of identifying coal mine drilling process conditions. First, the energy levels of coal mine drilling operating states are divided. By introducing a process capability index and using drilling rate as the calculation variable, state level labels are added to the coal mine drilling operating data. Actual data from a certain well site is used for calibration; some data is shown in Table 2.

[0112] Table 2 Classification of Coal Mine Drilling Operation Status Levels

[0113]

[0114]

[0115] (2) Parameter selection of coal mine drilling process condition identification model

[0116] The coal mine drilling data was filtered and normalized to obtain standardized coal mine drilling data. Feature extraction was performed using the SVD method to obtain the singular values ​​of each variable (see Table 3), which were then sorted by magnitude. The inflection point of the singular value fitting curve was used as the basis for threshold setting. Figure 1 It can be seen that a fixed threshold value of 3 is set for selecting feature variables. The results show that the singular vectors of three variables—feed pressure, torque, and power head displacement—are selected. Therefore, the singular vectors of these three parameters are chosen as the input parameters for the working condition identification model.

[0117] Table 3 Singular Values ​​of Variables

[0118] variable Feed pressure Torque Power head displacement rotational speed Drilling speed Singularity 11.1791 1.9043 0.0394 0.0204 0.0045

[0119] (3) Implementation of the state perception model for coal mine drilling process

[0120] The model's input consists of the singular vectors of the three coal mine drilling operation parameters selected in step (2): feed pressure, torque, and power head displacement. The model's output is the level (i.e., working condition) of the coal mine drilling operation status defined in step (1). The model is implemented using the GBDT method, and the model's identification results are shown in Table 4 below.

[0121] Table 4. Confusion Matrix for Identifying Coal Mine Drilling Conditions

[0122]

[0123] This model selected 112 sets of data from a certain well site, using the first 70% as the training set and the last 30% as the test set. The confusion matrix of the test results is shown in Table 4, and the final recognition results are shown below. Figure 4 .Depend on Figure 4 It can be seen that the model has an accuracy rate of 93.3% for the G1 level, 91.6% for the G2 level, and 83.3% for the G3 level, which verifies the effectiveness of the model to a certain extent.

[0124] The beneficial effects of this invention are: based on the drilling data of the coal mine drilling process, a method for classifying the coal mine drilling operation status level is defined by introducing a process capability index. Through the established coal mine drilling process condition identification model, the selected drilling parameter singular vector is input, and the model output is the coal mine drilling process operation status level (i.e., working condition). This invention lays the foundation for research on state monitoring, optimization, and control of the coal mine drilling process, and has practicality and applicability.

[0125] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for identifying working conditions during underground drilling in coal mines based on gradient boosting decision-making, characterized by comprising the following steps: Step 1: Classification of Operating Status Levels in Coal Mine Drilling Process: Introducing Process Capability Index With drilling speed To calculate the variables, the operational status level of the coal mine drilling process is obtained. Establish a state level classification corresponding to the process capability index, namely, three working conditions: normal, conservative, and excessive. Step 1 specifically includes: 1.1 Based on the operating status of the coal mine drilling process, analyze the changes in drilling process operating parameters and obtain the distribution of drilling speed, i.e. With the target drilling rate The corresponding process capability index value is obtained by using drilling rate as the calculation variable; the Taguchi process capability index calculation formula used is as follows: ； in and Do not adjust the distribution variance for the upper and lower tolerance limits of the selected sliding window. ; 1.2 The process capability index and corresponding performance level are as follows: When This is an overreaction; when This is the normal state; when This is a conservative state; Step 2: Parameter selection for coal mine drilling process condition identification model: After filtering and normalizing the drilling operation parameters, the singular value decomposition method is used to extract variable features of the drilling operation parameters. The inflection point of the singular value fitting curve is used as the basis for threshold setting. Drilling operation parameters with fixed thresholds are selected as input parameters for the drilling process condition identification model. Step 3: Implementation of the coal mine drilling process condition identification model: Using the singular vector of the input parameters obtained in Step 2 as the model input and the coal mine drilling process operation status level as the model output, the gradient boosting decision tree method is used to implement the model, and finally realize the identification of coal mine drilling process conditions. Step 3 specifically includes: 3.1 The singular vectors of feed pressure, torque, and power head displacement are used as inputs to the model; Based on the state level corresponding to the process capability index, state labels are added to the operational data as the output of the model. 3.2 The GBDT classification algorithm is used for model construction, and its implementation method is as follows: Define the training set as ,in m The number of samples is , and the final strong learner is . The loss function is: ； in Indicates whether it belongs to the first Class, 0 represents no, 1 represents yes; , The number of operational status level categories is 3; This represents the probability that a sample point belongs to each category.

2. The method for identifying the working conditions of underground coal mine drilling process based on gradient boosting decision-making according to claim 1, characterized in that, Step 2 specifically includes: 2.1 The maximum and minimum value normalization unified parameter format is adopted to filter and normalize the coal mine drilling process data to obtain standardized coal mine drilling data, and the singular value decomposition method is used to extract variable features; ； if I yes Matrix, then U yes unitary matrix; S yes A diagonal matrix in which the numbers arranged on the diagonal are real and non-negative. V The dimension is The unitary matrix, where represent V The conjugate transpose of; Singularity From and The non-zero square roots of the obtained eigenvalues ​​yielded two singular vectors; U and V The columns are called the left singular vector and the right singular vector, respectively; the diagonal matrix S The arrangement of singular values ​​is: ； in ; For matrix I The singular values; Singular values ​​are filtered based on a fixed threshold; non-zero singular values ​​are rearranged and a fixed number are selected from largest to smallest. i singular values, The inflection point of the singular value fitting curve is used as the basis for threshold setting. The coal mine drilling parameters are used as the input parameters of the model; 2.2 Based on the above judgment criteria, the drilling characteristic parameters of three features in coal mine drilling parameters—feed pressure, torque, and power head displacement—are selected as the input parameters of the model.

3. The method for identifying the working conditions of underground coal mine drilling process based on gradient boosting decision-making according to claim 1 or 2, characterized in that, The specific steps for building a model using the GBDT classification algorithm are as follows: Step 3.1: Initialize the weak learner; ； Step 3.2: In the first... t During the next iteration: Step 3.2-1: Calculate the probability that a sample point belongs to each category; ； Step 3.2-2: For each category k = 1, 2, …,K; Step 3.2-2-1: Calculate the negative gradient error; ； Step 3.2-2-2: [The text appears to be incomplete and contains several grammatical errors. A more Using the training data, the fitted data is obtained as follows: t A regression tree, For the corresponding leaf node region, j = 1, 2, …, J , J Given the number of leaf nodes, calculate the best-fit value for each leaf node region; ； Step 3.2-2-3: Update the strong learner; ； Step 3.3: The iteration is complete, and the final strong learner is obtained; ； The final result It can be used to calculate the number of... k The probability of the class corresponding to the class; ； ； For the final output category, When the actual value is k’ The prediction is the first k The joint cost of the classes, i.e., the class with the highest probability, is the output class.

Images