Intelligent optimization method for rotation stable inclination parameters for pneumatic directional drilling

Abstract

The invention relates to the field of underground coal mine drilling, in particular to an intelligent optimization method for rotation stable inclination parameters for pneumatic directional drilling, which comprises the following steps of: preprocessing acquired actual drilling data, complementing missing values through an average number filling method, and perfecting a drilling data set; based on the complemented data set, a mutual information analysis method is adopted to analyze the correlation between the characteristic quantity and the target value, and the final characteristic quantity is determined; fitting parameters of the drilling speed prediction model, the wind pressure prediction model and the pulling pressure prediction model by adopting a feedforward neural network method; and finally, forming a multi-objective optimization problem by taking the obtained drilling speed model, the wind pressure model and the pulling pressure model as objective functions and taking upper and lower bounds of the feeding pressure, the wind volume and the rotation pressure as constraint conditions, and solving by adopting an NSGA-II algorithm to obtain optimal operation parameters. The drilling efficiency can be improved, and the safety and economical efficiency of operation are enhanced.

Description

Technical Field

[0001] This invention relates to the field of underground drilling in coal mines, and in particular to an intelligent optimization method for slewing stability parameters in pneumatic directional drilling. Background Technology

[0002] In underground coal mine drilling operations, the rotation and stabilization process is a crucial step, and the optimization of operating parameters plays a decisive role in the smooth and efficient completion of the entire drilling operation. Increasing the drilling speed means drilling a greater distance per unit time, directly shortening the drilling cycle, which significantly promotes improved coal resource extraction efficiency. Simultaneously, reducing air pressure and pull-out pressure effectively reduces wear on drilling equipment during operation, lowering the risk of stuck drill bits due to increased friction between the drill bit and the borehole wall caused by excessive pressure. This ensures drilling safety, reduces downtime and economic losses due to equipment failure or accidents, and is of great significance for improving overall drilling efficiency and safety.

[0003] Traditional optimization studies for addressing borehole parameter optimization in coal mines often focus on multi-objective optimization of drilling speed and drill bit wear. However, in actual engineering, drill bit wear is influenced by multiple factors such as coal seam lithology and drilling technology, leading to difficulties in quantification and high real-time monitoring costs, resulting in insufficient adaptability of optimization models to field conditions. Furthermore, traditional methods only focus on the local synergy between drilling speed and wear, failing to correlate with real-time quantifiable parameters such as air pressure and pull-out pressure during pneumatic directional drilling. Unstable air pressure directly affects cuttings removal efficiency and drill string cooling, while unreasonable pull-out pressure may trigger the risk of stuck drill bit. These parameters, however, can be acquired and quantified in real time through sensors. Summary of the Invention

[0004] To address the impact of operating parameters on drilling performance during pneumatic directional drilling, this invention provides an intelligent optimization method for rotation and inclination stabilization parameters in pneumatic directional drilling. Focusing on the rotation and inclination stabilization stage of pneumatic directional drilling, it abandons the traditional approach relying on difficult-to-quantify wear indicators. Instead, it constructs a multi-objective optimization system using parameters that can be monitored in real time, such as drilling speed, air pressure, and pull-out pressure. By establishing a dynamic coupling model and optimization algorithm between parameters, it achieves a synergistic improvement in drilling efficiency and safety in complex formations, overcoming the limitations of traditional methods in terms of insufficient optimization accuracy and engineering adaptability due to the difficulty in quantifying indicators. This method mainly includes: S1: Preprocess the collected actual borehole data in the coal mine, and use the mean imputation method to fill in missing values ​​and improve the borehole dataset; S2: Based on the completed dataset, mutual information analysis is used to analyze the correlation between feature quantities and target values ​​to determine the final feature quantities; the feature quantities include feed pressure, air volume, slewing pressure, hole depth, coal and rock hardness, and slag return amount, and the target values ​​include drilling speed, air pressure, and pull-out pressure; S3: A feedforward neural network method is used to describe the relationship between the final feature quantity and the target value. A multi-output neural network target model is constructed. The multi-output neural network target model includes a drilling speed prediction model, a wind pressure prediction model, and a pull-out pressure prediction model. By adjusting the parameters of the feedforward neural network, multi-output neural network target models under different parameters are obtained. The root mean square error is used as the fitting performance index to obtain the optimal target model. The model input is the final feature quantity, and the model output is the target value. S4: The optimal multi-output neural network target model is used as the objective function, with the optimization objectives of maximizing drilling speed, minimizing air pressure, and minimizing pull-out pressure, and the upper and lower bounds of feed pressure, air volume, and slewing pressure as constraints, thus forming a multi-objective optimization problem. S5: The NSGA-II algorithm is used to solve the multi-objective optimization problem. The initial population size, number of genetic iterations and mutation rate are set. When the termination condition is met, the three-dimensional Pareto solution set is obtained. The optimal operating parameters in the slewing and stabilization process are obtained based on the linear weight method. The optimal operating parameters refer to maximizing the drilling speed and minimizing the air pressure and pull-out pressure.

[0005] A computer device includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-described method.

[0006] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above method.

[0007] A computer program product includes a computer program or instructions that, when executed by a processor, implement the steps of the above-described method.

[0008] The beneficial effects of the technical solution provided by this invention are as follows: This invention, through data collection and preprocessing, uses the mean imputation method to handle missing values, ensuring data integrity and accuracy. It then deeply analyzes the correlation between characteristic quantities and target values, and after screening, obtains key characteristic quantities. The constructed drilling speed model, air pressure model, and pull-out pressure model can accurately reflect the complex relationships and changing patterns among various parameters during the rotation and stabilization process in underground coal mines. Based on the constructed models, trajectory tracking control and multi-objective optimization are carried out. The objective function, decision variables, and constraints are rationally determined. Using the NSGA-II strategy, the optimal combination of operating parameters is successfully obtained, which can effectively improve drilling speed, reduce in-hole air pressure and pull-out pressure, thereby improving drilling efficiency and enhancing the safety and economy of the operation. Verification using actual underground coal mine drilling operation data shows that the obtained optimized parameters exhibit excellent performance, shortening the drilling cycle and enhancing operational safety and stability. This provides effective technical support and optimization solutions for underground coal mine drilling engineering, and powerfully promotes the progress and development of coal mine tunnel drilling technology. Attached Figure Description

[0009] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 This is a flowchart of an intelligent optimization method for slewing stabilization parameters for pneumatic directional drilling, as described in an embodiment of the present invention. Figure 2 This is a heat map showing the mutual information dependence of the 60# rotational stabilization process in this embodiment of the invention. Figure 3 This is a heat map showing the mutual information dependence of the 65# rotational stabilization process in this embodiment of the invention. Figure 4 This is a heat map showing the mutual information dependence of the 95# rotational stabilization process in this embodiment of the invention. Figure 5 This is a heat map showing the mutual information dependence of the 100# rotational stabilization process in this embodiment of the invention. Figure 6 This is a heatmap showing the average mutual information dependence of the slewing and tilting process in this embodiment of the invention. Figure 7 This is the fitting result of the feedforward neural network model with hidden layer parameters [30, 15, 30] in this embodiment of the invention; Figure 8 This is the fitting result of the feedforward neural network model with hidden layer parameters [20, 20, 30] in this embodiment of the invention; Figure 9 This is the fitting result of the feedforward neural network model with hidden layer parameters [30, 30] in the embodiment of the present invention; Figure 10 This is the solution set of operation parameters based on the NSGA-II algorithm in this embodiment of the invention; Figure 11 This is the multi-objective optimization result based on the NSGA-II algorithm in the embodiments of the present invention. Detailed Implementation

[0010] 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.

[0011] Example 1 Please refer to Figure 1 , Figure 1 This is a flowchart of an intelligent optimization method for slewing stabilization parameters in pneumatic directional drilling, as described in an embodiment of the present invention, specifically including: S1: Preprocess the collected actual borehole data in the coal mine, and use the average filling method to complete the borehole data to obtain a complete borehole dataset.

[0012] The data used in this embodiment is based on actual borehole data from a mine in Hancheng. However, this data has some missing data. Therefore, the mean imputation method is used to impute the missing values. The mean imputation method is a statistical imputation technique used to handle missing values ​​in numerical data. Its basic idea is to replace the missing values ​​with the arithmetic mean of all non-missing values ​​of the feature.

[0013] Suppose there is a feature column X This includes some missing values ​​(usually represented as NaN or other specific symbols). First, we need to calculate the mean of all non-missing values ​​in this feature column. Let... X ={ x 1, x 2, ... , x n} is a containing n The feature column of each observation contains values ​​that may be missing. Only the non-missing values ​​are considered when calculating the mean. Let... S S is the set of all non-missing values, i.e., S = { x i | x i , ... ,ϵ X}

[0014] The mean of the non-missing values It can be calculated using the following formula:

[0015] in, This represents the summation of all non-missing values. This indicates the number of non-missing values.

[0016] S2: The correlation between characteristic quantities (feed pressure, air volume and slewing pressure, hole depth, coal and rock hardness and slag return) and target values ​​(drilling speed, air pressure and pull-out pressure) is calculated using mutual information analysis. Characteristic quantities with high correlation are retained. The correlation analysis process is as follows: A complete dataset was obtained through preprocessing of the borehole data. Based on this dataset, mutual information analysis was used to analyze the correlation between characteristic quantities (feed pressure, air volume and slewing pressure, hole depth, and coal / rock hardness) and target values ​​(drilling rate, air pressure, and pull-out pressure). Mutual information (MI) is a measure of the dependency between two random variables; it can capture not only linear but also nonlinear relationships. Mutual information can be used for correlation analysis between continuous and discrete variables.

[0017]

[0018] in, and Let there be two random variables. yes and The joint probability distribution function, and They are and The marginal probability distribution function, express and The higher the correlation value, the stronger the correlation.

[0019] This invention uses data from four holes (60#, 65#, 95#, and 100#) for mutual information analysis. The results of the mutual information analysis are as follows: Figure 2-6 As shown, Figure 2 The heatmap shows the mutual information dependence of the 60# rotational steady-state tilting process. Figure 3 The heatmap shows the mutual information dependence of the 65# rotational steady-state tilting process. Figure 4 The heatmap shows the mutual information dependence of the 95# rotational steady-state tilting process. Figure 5 The heatmap shows the mutual information dependence of the 100# rotational steady-state tilting process. Figure 6This is a heatmap showing the average mutual information dependency during the rotary stabilization process. The mutual information analysis results show that although the correlations between characteristic quantities (feed pressure, air volume and rotary pressure, hole depth, coal and rock hardness, and slag return amount) and target values ​​(drilling speed, air pressure, and pull-out pressure) vary across different holes, they generally exhibit some degree of correlation. The correlation coefficient is used to assess the degree of correlation between variables, and the average correlation coefficient is mostly higher than 0.5 (according to the mutual information correlation coefficient rule, a coefficient below 0.5 indicates a weak correlation). In the underground rotary stabilization process in coal mines, only the slag return amount shows a poor correlation with the target quantity, with an average correlation coefficient below 0.5. Therefore, based on the principle of feature extraction, the slag return amount is removed, and only feed pressure, air volume and rotary pressure, hole depth, and coal and rock hardness are retained as the final characteristic quantities. It should be noted that correlation describes the existence of a statistical dependency between two variables, a qualitative description; while the correlation coefficient is a quantitative indicator used to measure the strength and direction of this relationship. In this invention, the mutual information correlation coefficient is used to measure the degree of nonlinear dependence between variables, and is therefore more suitable for describing the influence relationship between complex variables during the drilling process.

[0020] S3: Using the feature values ​​as model inputs and the target values ​​as model outputs, a multi-output neural network target model (drilling speed prediction model, wind pressure prediction model, and pull-out pressure prediction model) is constructed using the feedforward neural network method in basic deep learning models, and then fitted. By adjusting the neural network parameters, multi-output neural network target models under different parameters are obtained. The root mean square error is used as the fitting performance index to select the multi-output neural network target model with the best fitting effect. During the slewing and stabilization process, drilling speed, air pressure, and pull-out pressure are affected by multiple factors, mainly including feed pressure, air volume and slewing pressure, hole depth, and coal and rock hardness. The interaction between these factors is crucial for optimizing the slewing and stabilization process.

[0021] To establish a model that comprehensively considers the above factors, a feedforward neural network is used for modeling. As a data-driven nonlinear modeling tool, the feedforward neural network's multi-layered hidden layer structure can automatically capture the complex mapping relationship between input variables (feed pressure, air volume, borehole depth, coal and rock hardness, etc.) and output targets (drilling speed, air pressure, pull-out pressure) through adaptive learning, without requiring a predefined specific function form. The nonlinear transformation characteristics of the activation function in the network can extract abstract features of the data layer by layer, effectively handling nonlinear interactions between variables, and is particularly suitable for engineering scenarios with complex mechanisms and unstandardized relationships. The specific modeling process is as follows: First, data reading and feature extraction are performed, reading data from multiple borehole files and extracting independent variables. X and dependent variable YThe borehole data is then divided into training and testing sets, and features are standardized. To balance model fitting time and accuracy, a random search method is used to configure the activation and training functions. The activation function determines the shape of the fitting function (non-linear modeling capability), while the training function determines how the fitting function is optimized (training efficiency and convergence speed). The fitting function itself is the target criterion for measuring prediction error. In this embodiment, the activation function options are set to 'logsig', 'tansig', 'purelin', etc., and the training function options are set to 'trainlm', 'trainscg', 'trainbr', etc. The activation and training functions respectively determine the model's expressive power and parameter optimization method, both serving the process of minimizing the fitting target function. Finally, a multi-output neural network target model is obtained through fitting.

[0022] in, X =[ x 1, x 2, x 3, x 4, x 5] T , x 1, x 2, x 3, x 4, x 5 represents feed pressure, air volume, rotational pressure, hole depth, and coal / rock hardness, respectively. Y =[ y 1, y 2, y 3] T , y 1, y 2, y 3 are the drilling speed prediction model, the air pressure prediction model, and the pull-out pressure prediction model, respectively; in the hidden layer, z (l) For the first l layer( l The result of a linear combination of (=1,2,3), W (l) For the first l The weight matrix of the layer, b (l) For the first l Layer bias vector, a (l) For the first l The activation value of the layer; in the output layer, z out The result is a linear combination of the output layer. W outThe weight matrix of the output layer. b out This is the bias vector for the output layer. For the neural network to the first i The predicted output vector for each input sample. Indicates the neural network's first... The activation function output of a layer is an intermediate result after non-linear processing, and is usually used to enhance the model's fitting ability. This represents the result of the activation function processing in the output layer, and is the final predicted value of the model.

[0023] After obtaining the target model of the multi-output neural network, it is necessary to discuss the model fitting effect under different hidden layer parameters, so as to select the best hidden layer parameters. Under different hidden layer parameters, the predicted values ​​of drilling speed, air pressure, and pull-out pressure obtained by the feedforward neural network are compared with the actual values ​​measured on site, such as... Figure 7-9 As shown, Figure 7 The results are fitting results for a feedforward neural network model with hidden layer parameters [30, 15, 30]. Figure 8 The fitting results are based on a feedforward neural network model with hidden layer parameters [20, 20, 30]. Figure 9 The fitting results are based on a feedforward neural network model with hidden layer parameters [30, 30].

[0024] After obtaining the fitting results of the drilling speed prediction model, air pressure prediction model, and pull-out pressure prediction model, it is necessary to evaluate the performance of the fitted models, using the root mean square error (RMSE) as the evaluation index for model fitting. The formula for calculating the root mean square error (RMSE) is:

[0025] in n It is the sample size. It is the first The actual value of each sample It is the first The predicted value for each sample.

[0026] Based on the performance evaluation index of the feedforward neural network model fitting, the model with the smallest root mean square error is selected as the final multi-output neural network target model for constructing the objective function.

[0027] S4: Taking the fastest drilling speed, minimum air pressure, and minimum pull-out pressure as objectives, the optimal multi-output neural network objective model is used as the objective function, highly correlated features are used as decision variables, and constraints are set based on actual engineering experience to construct a multi-objective optimization problem. In the process of stabilizing the borehole in underground coal mines, multi-objective optimization of drilling speed, air pressure, and pull-out pressure is crucial. Optimizing drilling speed can improve drilling efficiency and shorten the drilling cycle, thereby saving time and costs; while optimizing air pressure and pull-out pressure can reduce drill bit wear and failure risks, avoid problems such as stuck drill bit caused by excessive air pressure and pull-out pressure, and thus improve the safety and reliability of drilling operations.

[0028] (1) Objective function In this section, facing the optimization problem, the selected objective functions are the drilling speed prediction model, the air pressure prediction model, and the pull-out pressure prediction model. The objective functions are established with the fastest drilling speed and the lowest air pressure and pull-out pressure as the optimization objectives:

[0029] in, x 1, x 2, x 3, x 4, x 5 correspond to characteristic quantities (feed pressure, air volume, rotation pressure, hole depth, coal and rock hardness), while These correspond to the target quantities (drilling speed, air volume, and pull-out pressure), respectively.

[0030] (2) Decision variables In the slewing and stabilization process of underground coal mines, feed pressure is a core parameter directly controlled by engineers. It directly affects drilling speed, air pressure, and pull-out pressure by altering the force exerted by the drill bit on the rock. While higher feed pressure enhances the drill bit's rock-breaking ability and increases drilling speed, it also simultaneously increases pull-out pressure, exacerbating equipment wear risks. Using it as a decision variable allows for a balance between drilling speed improvement and equipment safety, optimizing drilling efficiency. Air volume is another key controllable parameter, its core function being to influence the efficiency of gas carrying rock cuttings. Using air volume as a decision variable allows for the coordinated optimization of air pressure and pull-out pressure. Slewing pressure, as a directly controllable parameter driving the drill bit's rotation, determines the frictional shear force at the drill bit's contact surface with the rock, directly affecting rock-breaking efficiency and drill string rotation stability. Reasonable adjustment of slewing pressure can improve rock-breaking efficiency, stabilize the drilling trajectory, and prevent excessive drill string wear due to excessive pressure. In addition, hole depth, as an uncontrollable characteristic that increases naturally with drilling time, together with coal and rock hardness (another uncontrollable characteristic) which reflects the physical properties of the rock, affects drilling speed, air pressure and pull-out pressure. Hard coal seams require higher feed pressure to achieve the same drilling speed and will increase pull-out pressure; soft coal seams have lower pressure requirements, but the risk of rock cuttings blockage must be prevented. In summary, feed pressure, air volume, slewing pressure, hole depth, and coal / rock hardness (uncontrollable characteristics) were chosen as decision variables because of their direct correlation with drilling speed, air pressure, and pull-out pressure. By adjusting the controllable parameters, drilling speed can be maximized while air pressure and pull-out pressure are minimized. Combined with the evaluation of uncontrollable factors, a variable relationship model is finally established through fitting experimental data to achieve systematic optimization of borehole performance.

[0031] (3) Constraints In the process of slewing and stabilizing the shaft in underground coal mines, constraints are crucial to ensuring that decision variables such as feed pressure, air volume, slewing pressure, borehole depth, and coal / rock hardness operate within safe and effective ranges during optimization. These constraints are based on limitations imposed by equipment capacity, coal seam characteristics, safety requirements, and operating procedures, aiming to guarantee the safety, stability, and efficiency of slewing and stabilizing the shaft.

[0032] Feed pressure, airflow, and rotary pressure are affected by factors such as drill bit structure and drilling rig power. Each drill bit model has specified minimum and maximum airflow and rotary pressure. , To provide the upper and lower bounds of the pressure, The upper and lower limits of air volume, This represents the upper and lower limits of the rotational pressure.

[0033] The overall multi-objective optimization problem for underground coal mine drilling parameters is as follows:

[0034] in, For the upper and lower bounds of the variable, To provide the upper and lower bounds of the pressure, The upper and lower limits of air volume, This represents the upper and lower limits of the rotational pressure.

[0035] The hole depth is calculated based on the number of drill rods and the length of each drill rod, and its value is updated progressively during the optimization process. The coal and rock hardness is updated based on the actual hole depth.

[0036] S5: Based on the multi-objective optimization problem, the NSGA-II algorithm is used to solve the problem. The initial population size, number of genetic iterations and mutation rate are set to obtain the three-dimensional Pareto set. Then, the optimal operating parameters are selected based on the weight method. Finally, the drilling speed, air pressure and pull-out pressure are optimized through the optimal operating parameters. The NSGA-II optimization algorithm was used to optimize the objective parameters based on the drilling speed prediction model, air pressure prediction model, and pull-out pressure prediction model. First, the Pareto set was obtained, and then the multi-objective optimization result was obtained based on the weighted method. The final optimization result is as follows: Figure 10-11 As shown, Figure 10 For the solution set of operation parameters based on the NSGA-II algorithm, Figure 11 The results show the multi-objective optimization based on the NSGA-II algorithm. As can be seen from the figure, through optimization, the drilling speed increased from an average of 1.39 m / min to 1.73 m / min, an increase of 0.34 m / min (approximately 24.5%); the air pressure decreased from an average of 0.97 MPa to 0.79 MPa, a reduction of 0.18 MPa (approximately 18.6%); and the pull-out pressure decreased from an average of 3.83 MPa to 3.62 MPa, a reduction of 0.21 MPa (approximately 5.5%).

[0037] Example 2 A computer device includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-described method.

[0038] Example 3 A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above method.

[0039] Example 4 A computer program product includes a computer program or instructions that, when executed by a processor, implement the steps of the above-described method.

[0040] 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 intelligent optimization of slewing and tilt stabilization operation parameters for pneumatic directional drilling, characterized in that, S1: Preprocess the collected actual borehole data in the coal mine, and use the mean imputation method to fill in missing values ​​and improve the borehole dataset; S2: Based on the completed dataset, mutual information analysis is used to analyze the correlation between feature quantities and target values ​​to determine the final feature quantities; The characteristic quantities include feed pressure, air volume, rotation pressure, hole depth, coal and rock hardness, and slag return amount; the target values ​​include drilling speed, air pressure, and pull-out pressure; and the final characteristic quantities include feed pressure, air volume, rotation pressure, hole depth, and coal and rock hardness. S3: A feedforward neural network method is used to describe the relationship between the final feature quantity and the target value. A multi-output neural network target model is constructed. The multi-output neural network target model includes a drilling speed prediction model, a wind pressure prediction model, and a pull-out pressure prediction model. By adjusting the parameters of the feedforward neural network, multi-output neural network target models under different parameters are obtained. The root mean square error is used as the fitting performance index to obtain the optimal target model. The model input is the final feature quantity, and the model output is the target value. S4: The optimal multi-output neural network target model is used as the objective function, with the optimization objectives of maximizing drilling speed, minimizing air pressure, and minimizing pull-out pressure, and the upper and lower bounds of feed pressure, air volume, and slewing pressure as constraints, thus forming a multi-objective optimization problem. S5: The NSGA-II algorithm is used to solve the multi-objective optimization problem. The initial population size, number of genetic iterations and mutation rate are set. When the termination condition is met, the three-dimensional Pareto solution set is obtained. The optimal operating parameters in the slewing and stabilization process are obtained based on the linear weight method. The optimal operating parameters refer to maximizing the drilling speed and minimizing the air pressure and pull-out pressure.

2. The intelligent optimization method for slewing and tilt stabilization operation parameters for pneumatic directional drilling according to claim 1, characterized in that, In S1, the process of imputing missing values ​​using the mean imputation method is as follows: set up X ={ x 1, x 2, ... , x n } is a containing n Given a feature column of observations, some of which may be missing values, the mean is calculated considering only the non-missing values. Let... S S is the set of all non-missing values, i.e., S = { x i | x i , ... ,ϵ X }, then the mean of the non-missing values Calculated using the following formula: in, This represents the summation over all non-missing values. Indicates the number of non-missing values; Replace missing values ​​with the mean to improve the borehole data.

3. The intelligent optimization method for slewing and tilt stabilization operation parameters for pneumatic directional drilling according to claim 1, characterized in that, In S2, the formula for calculating correlation is: in, and Let there be two random variables. yes and The joint probability distribution function, and They are and The marginal probability distribution function, express and The correlation.

4. The intelligent optimization method for slewing and tilt stabilization operation parameters for pneumatic directional drilling according to claim 1, characterized in that, In S3, the modeling process is as follows: (1) Read data from multiple borehole files and extract the independent variables. X and dependent variable Y ; (2) Divide the borehole dataset into a training set and a test set, and perform feature standardization; (3) In order to balance model fitting time and fitting accuracy, a random search method is used to configure the activation function and training function; (4) Obtain the target model of the multi-output neural network through fitting: in, X =[ x 1, x 2, x 3, x 4, x 5] T , x 1, x 2, x 3, x 4, x 5 represents feed pressure, air volume, rotational pressure, hole depth, and coal / rock hardness, respectively. Y =[ y 1, y 2, y 3] T , y 1, y 2, y 3 are the drilling speed prediction model, the air pressure prediction model, and the pull-out pressure prediction model, respectively. z (l) For the hidden layer l The result of linear combination of layers, W (l) For the hidden layer l The weight matrix of the layer, b (l) For the hidden layer l The layer's bias vector, a (l) For the hidden layer l The activation value of the layer, l =1,2,3; z out The result is a linear combination of the output layer. W out The weight matrix of the output layer. b out This is the bias vector for the output layer. For the neural network to the first i The predicted output vector for each input sample. Indicates the neural network's first... The activation function output of the layer, This represents the result of the activation function processing in the output layer, and is the final predicted value of the model.

5. The intelligent optimization method for slewing and tilt stabilization operation parameters for pneumatic directional drilling according to claim 4, characterized in that, In S4, an objective function is constructed based on the requirements of the slewing and tilt stabilization process to ensure that the wind pressure and pull-out pressure are minimized while maximizing the drilling speed. The objective function is: in, These correspond to drilling speed, air volume, and pulling pressure, respectively.

6. The intelligent optimization method for slewing and tilt stabilization operation parameters for pneumatic directional drilling according to claim 5, characterized in that, In S4, the multi-objective optimization problem is: in, For the upper and lower bounds of the variable, To provide the upper and lower bounds of the pressure, The upper and lower limits of air volume, This represents the upper and lower limits of the rotational pressure.

7. The intelligent optimization method for slewing and tilt stabilization operation parameters for pneumatic directional drilling according to claim 1, characterized in that, In S5, the initial population, number of genetic iterations, and mutation rate are first initialized using the NSGA-II algorithm, where individuals represent combinations of drilling parameters. Then, individuals are selected through non-dominated sorting and crowding calculation. After selection, crossover, and mutation operations, offspring populations are generated and merged. This process is repeated iteratively to obtain a three-dimensional Pareto set. Then, the optimal operating parameters are selected based on the weighting method. Finally, the target model is optimized using the optimal operating parameters to obtain the objective function, thus solving the multi-objective optimization problem, maximizing drilling speed, minimizing wind pressure and pull-out pressure, and improving drilling efficiency.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes a computer program to implement the steps of the intelligent optimization method for slewing and tilt stabilization operation parameters for pneumatic directional drilling as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, implements the steps of the intelligent optimization method for slewing and tilt stabilization operating parameters for pneumatic directional drilling as described in any one of claims 1-7.

10. A computer program product, characterized in that, It includes a computer program or instructions that, when executed by a processor, implement the steps of the intelligent optimization method for slewing and tilt stabilization operation parameters for pneumatic directional drilling as described in any one of claims 1-7.

Images