A rock porosity prediction method based on small sample while-drilling data inversion

Through the method of combining multiple noise reduction processing and neural networks with genetic algorithms, a rock porosity convolutional neural network model is constructed using small sample drilling data, solving the problem of rock porosity inversion and achieving rapid and accurate in tunnel engineering Porosity prediction reduces the dependence of laboratory testing.

CN118568629BActive Publication Date: 2024-11-15CHINA UNIV OF MINING & TECH
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202410696914.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-31
Publication Date
2024-11-15
Estimated Expiration
2044-05-31

AI Technical Summary

Technical Problem

The prior art is difficult to use small sample drilling data to quickly and accurately invert rock porosity, especially in shallow tunnel engineering. Traditional measurement methods are economically costly and cumbersome, and the existing literature lacks pores for tunnel surrounding rocks. inversion prediction method of degree.

Method used

Using multiple noise reduction processing and neural network combined with genetic algorithm, the drilling parameters of the drilling rig are obtained, and the prediction model of rock porosity convolutional neural network is constructed, and data augmentation and model training is used for time-frequency domain feature image library to achieve rock porosity fast quantitative prediction.

Benefits of technology

Effectively solve the problem of small data volume and susceptible to external interference during data acquisition while drilling, improve the inversion accuracy of rock porosity, realize fast and accurate porosity prediction on site, and reduce the dependence on laboratory testing. .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118568629B_ABST
    Figure CN118568629B_ABST
Patent Text Reader

Abstract

The invention discloses a rock porosity prediction method based on small sample downhole data inversion. Downhole parameters of a drilling rig in a laboratory or on-site drilling process are collected, including torque M, thrust F, rotation speed N, thrust speed V, drill pipe amplitude A, and vibration acceleration a parameter; different types of downhole parameter data after multiple noise reduction processes are input into a BP‑GA model for discrete point elimination, data enhancement, and iterative calculation to form a new downhole data set; time-frequency domain characteristic graphs of downhole parameters of rocks with different porosities are input into a VG‑CNN convolutional neural network prediction model, and an inversion model between the time-frequency domain characteristics of downhole parameters and porosity is obtained through training and learning, and finally the inversion prediction of rock porosity is realized according to the on-site real-time downhole parameters; the method can perform fast, accurate, and quantitative inversion prediction of rock porosity based on a small amount of borehole downhole data.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to an inversion prediction method for underground engineering rock porosity, and in particular to a rock porosity prediction method based on small sample while-drilling data inversion. Background Art

[0002] Porosity is an important indicator to characterize the physical and mechanical parameters of rocks, such as strength, elastic modulus, and permeability. Rapid determination of rock porosity in underground tunnel projects is of great significance for the selection of support methods and the determination of support parameters. The traditional method of measuring rock porosity requires processing the cored rock into test samples. Conduct laboratory test analysis. In the fields of oil drilling, the porosity of oil and gas reservoirs can be finely characterized and quantitatively inverted by analyzing logging information. However, for shallow tunnel projects (such as subways, tunnels, underground tunnels, etc.), the economic cost of logging and logging is high and the operation is cumbersome. Therefore, it is necessary to propose a quantitative, fast, and economical in-situ measurement method for rock porosity.

[0003] At present, the drilling test technology has gradually become a hot topic in the field of intelligent exploration of geotechnical engineering. The drilling test technology can use the characteristics of the drilling parameters change during the drilling process to reflect the mechanical properties of the surrounding rock, and identify the lithology and structural characteristics of the formation. It can effectively make up for the time lag of the traditional measurement method and does not affect the on-site engineering construction. It is a convenient and fast in-situ measurement method. The relevant research results in the field of drilling test have preliminarily verified the feasibility of inverting rock strength and identifying rock structure surfaces based on drilling parameters. However, due to the complex process of drilling and rock cutting; at the same time, the on-site acquisition process of drilling parameters has the characteristics of small amount of effective data and easy interference from the external environment, which leads to great difficulties in using drilling parameters to invert and predict rock porosity. There are few methods for inversion prediction of tunnel surrounding rock porosity in the existing literature, and there is no relevant report on how to use small sample drilling data to accurately invert and predict rock porosity. Therefore, it is necessary to establish a scientific and reasonable method for inversion prediction of rock porosity while drilling. . Summary of the invention

[0004] In view of the above-mentioned technical deficiencies, the purpose of the present invention is to provide a rock porosity prediction method based on small sample downhole data inversion, which can use small sample downhole data in the laboratory or on-site drilling process to perform rapid and quantitative inversion prediction of rock porosity.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0006] The present invention provides a rock porosity prediction method based on small sample while-drilling data inversion, comprising the following steps:

[0007] Step S1: Collecting drilling information of the drilling rig during drilling in the laboratory or underground tunnel, and obtaining key drilling parameters such as torque M, thrust F, rotation speed N, thrust speed V, drill pipe amplitude A, and vibration acceleration a;

[0008] Step S2: performing multiple noise reduction processing on the different types of while-drilling parameters collected, inputting the while-drilling parameter values ​​after multiple noise reduction into the "BP-GA" model to obtain a new while-drilling parameter derived data set;

[0009] Step S3: analyzing the variation curves of different types of while-drilling parameters with the drilling time t in the new while-drilling parameter data set to obtain the time-frequency domain graph of the while-drilling parameters, and obtaining the time-frequency domain image library of different types of while-drilling parameters through the data enhancement function in the neural network;

[0010] Step S4: construct a rock porosity convolutional neural network prediction model, learn and train the time-frequency domain feature image library of drilling parameters during the drilling process of rocks with different porosities, determine the initial learning rate and attenuation coefficient of the model, and obtain the rock porosity prediction result.

[0011] Preferably, the implementation process of the multiple noise reduction methods for different types of drilling parameters in step S2 includes the following steps:

[0012] Step 1.1: Based on the raw data of different types of drilling parameters collected continuously, calculate the extreme point m of each type of drilling parameter raw data i , average value n i and the local amplitude a i ;

[0013]

[0014]

[0015] The local mean function a(t) is used to separate the original while-drilling data x(t), and then the envelope estimation function m(t) is used to demodulate the original while-drilling data x(t):

[0016] h(t)=x(t)-a(t)(3)

[0017]

[0018] Repeat the above operation until the original while-drilling data x(t) is reduced to a monotonic function or has ≤3 extreme points, then stop the iterative calculation. At this time, the original while-drilling data x(t) is decomposed into k decision components PF and a decomposition residual;

[0019] Step 1.2: Taking kurtosis H1, correlation coefficient H2 and root mean square value H3 as the selection criteria of PF components, the evaluation index V of PF components is ij It is expressed as:

[0020]

[0021] In the formula, j is 1, 2, 3, ..., k;

[0022] Here we define V ij >90% of the PF components are effective components;

[0023] Step 1.3: Use the wavelet denoising method to further denoise the filtered PF components, including wavelet denoising of the effective PF components of torque M, thrust F, rotation speed N, propulsion speed V, drill pipe amplitude A, and vibration acceleration a. The wavelet denoising function used for torque and rotation speed is the db3 wavelet denoising function, the wavelet denoising function used for thrust and propulsion speed is the coif3 wavelet denoising function, and the wavelet denoising function used for drill pipe amplitude and vibration acceleration is the bior3.3 wavelet denoising function;

[0024] Step 1.4: Input the multiple denoised LWD parameter data into the “BP-GA” model for iteration, crossover, and mutation to achieve data enhancement and data derivation.

[0025] Preferably, in step S2, a "BP-GA" model is established to solve the problem of small amount of sample data in the time-frequency domain graph caused by difficulty in collecting field data of the above six types of drilling parameters, and the specific method is as follows:

[0026] Step 2.1: Use the neural network BP to determine the collected while-drilling information samples and expected output values, and use the adaptive algorithm to determine the initial weight and threshold of the BP neural network corresponding to each type of while-drilling parameter, so as to eliminate unreasonable data;

[0027] Step 2.2: For the six types of drilling parameters after eliminating unreasonable data, genetic algorithm is used for iteration, crossover and data mutation. The crossover probability is 0.75 and the coefficient of variation is 0.031. The new drilling parameter data groups generated after encoding are M1, M2, ..., M6;

[0028] Step 2.3: Establish relationship curves between the newly generated while-drilling parameter data and the drilling time t, and then obtain time-frequency domain characteristic diagrams of different types of while-drilling parameters;

[0029] Step 2.4: Using the data enhancement function in the neural network, the time-frequency domain feature map of each type of drilling parameters is randomly flipped, rotated clockwise or counterclockwise, the brightness is changed, and the pixel value is increased to expand the time-frequency domain image library of the drilling parameters. When the number of time-frequency domain feature images corresponding to each type of drilling parameters is ≥1200, the iterative calculation of the neural network is stopped.

[0030] Preferably, in step S3, the method for acquiring time-frequency domain characteristic graphs of different types of drilling parameters comprises the following steps:

[0031] Step 3.1: Based on the drilling information processed by multiple noise reduction and “BP-GA” model, the variation curves of different types of drilling parameters with drilling time t are established, that is, the time domain characteristic diagram of the drilling parameters is obtained;

[0032] Step 3.2: Use discrete Fourier transform to transform the time domain characteristic diagram of the drilling parameters into the frequency domain characteristic diagram. The change equation is:

[0033]

[0034] Where Y(n) represents a finite-length discrete frequency sequence of length L, n = 0, 1, ..., L-1, Hz, and y(t) represents a discrete time sequence of length T;

[0035] Step 3.3: The time domain and frequency domain graphs of the drilling parameters are fused through short-time Fourier transform; that is, with time as the horizontal axis and the acquisition frequency and instantaneous energy change of the drilling parameters as the vertical axis, the time-frequency domain feature images of different types of drilling parameters are obtained, and the expanded time-frequency domain image library of the drilling parameters is input into the rock porosity convolutional neural network prediction model, namely the VG-CNN prediction model.

[0036] Preferably, in step S4, the specific prediction process of the rock porosity convolutional neural network prediction model is as follows:

[0037] Step 4.1 First, the time-frequency domain feature maps of the drilling information of rocks with different porosities are input into the first convolutional layer of the VG-CNN prediction model, which consists of 42 convolutional kernels of size 3×3;

[0038] Step 4.2 Secondly, the feature map after the first convolution is input into the first pooling layer of the VG-CNN prediction model, which consists of a 2×2 pooling window;

[0039] Step 4.3 Repeat steps 4.1 and 4.2 until the learning rate of the VG-CNN prediction model is ≤ 3×10 -5 , attenuation coefficient ≤ 1×10 -6 , stop model training; at this time, substitute the rock porosity prediction confusion matrix data into the following formula to verify the prediction accuracy of the VG-CNN prediction model for rock porosity.

[0040]

[0041] Where: P T represents a true example, N T represents a true counterexample, P Frepresents a false positive, N F Indicates a false counterexample.

[0042] The beneficial effects of the present invention are:

[0043] 1. This method obtains limited real-time downhole parameter data during the drilling process, and iteratively derives it through "neural network + genetic algorithm" after multiple noise reduction processes. The newly formed downhole data set has the characteristics of large data volume, high reliability and strong robustness, which effectively solves the technical problems of large external interference, small effective data volume and strong data discreteness in the data collection process of downhole experiments.

[0044] 2. The drilling parameters selected by this method are reasonable and scientific, which can truly reflect the rock porosity. By optimizing the structural parameters of the convolutional neural network, the inversion accuracy of the rock porosity is improved, and the lag caused by the previous quantitative evaluation of rock porosity needing to be sent to the laboratory for relevant tests on site is solved. This method can be applied to the prediction and evaluation of surrounding rock porosity in underground projects with changeable geological conditions and limited space, such as coal mine tunnels, subways, and tunnels. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 A schematic flow chart of a rock porosity prediction method based on small sample while-drilling data inversion provided by an embodiment of the present invention;

[0047] Figure 2 This is the structural diagram of the BP-GA intelligent algorithm derived model. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] like Figure 1 As shown, a rock porosity prediction method based on small sample while drilling data inversion includes the following steps:

[0050] Step S1: Collecting the drilling information of the drilling rig during the laboratory or on-site tunnel drilling process through sensors to obtain key drilling parameters such as torque M, thrust F, rotation speed N, thrust speed V, drill pipe amplitude A, and vibration acceleration a;

[0051] Step S2: multiple denoising is performed on the different types of collected while drilling parameters, and the while drilling parameter values ​​after multiple denoising are input into the "BP-GA" intelligent algorithm derivative model to obtain a new while drilling parameter data set;

[0052] Step S3: analyzing the variation curves of different types of while-drilling parameters with the drilling time t in the while-drilling parameter data set to obtain a time-frequency domain diagram of the while-drilling parameters, and obtaining a time-frequency domain image library of different types of while-drilling parameters through the data enhancement function in the neural network;

[0053] Step S4: construct a rock porosity convolutional neural network (VG-CNN) prediction model, learn and train the time-frequency domain feature image library of drilling parameters during the drilling process of rocks with different porosities, determine the initial learning rate and attenuation coefficient of the model, and obtain the rock porosity prediction result.

[0054] Here, laboratory tests are taken as an example to carry out rock porosity prediction based on drilling parameters inversion. First, rocks with different porosities such as sandstone, mudstone, limestone, granite, and marble are uniformly processed into 15cm×15cm×15cm cubic specimens, which are divided into 5 groups according to different lithologies, with 9 rock samples in each group, totaling 45 pieces. The drilling test experiment adopts the control variable method, that is, the drilling speed V and rotation speed N are kept constant during the drilling process, and the real-time data of other drilling parameters during the drilling process are monitored. The drilling speed is set to 60mm / min, the drill bit speed is set to 200r / min, and the drilling depth is uniformly preset to 120mm. After reaching the predetermined drilling depth, the drilling device automatically stops drilling.

[0055] First, the multiple noise reduction methods proposed in this patent are used to reduce the noise of the downhole parameters such as torque M, thrust F, rotation speed N, thrust speed V, drill pipe amplitude A, and vibration acceleration a collected by the sensor. Since the drilling speed V and rotation speed N are kept constant during the experiment, the focus is on noise reduction for torque M, thrust F, drill pipe amplitude A, and vibration acceleration a. The wavelet denoising function used for torque M is the db3 wavelet denoising function, the wavelet denoising function used for thrust is the coif3 wavelet denoising function, and the wavelet denoising function used for drill pipe amplitude A and vibration acceleration a is the bior3.3 wavelet denoising function; until each type of original downhole data x(t) is reduced to a monotonic function or has ≤3 extreme points, stop noise reduction. After noise reduction is completed, the data sets corresponding to each type of downhole parameter are N1, N2, ..., N6 respectively.

[0056] Attached Figure 2The model structure of the "BP-GA" intelligent algorithm is derived. The initial weight and screening threshold of the sample data are determined according to the results of the drilling experiment. The data range and average value are calculated between the same type of drilling parameters to propose data points with strong discreteness. After eliminating unreasonable data, the six types of drilling parameters are iterated, crossed and mutated using genetic algorithms. The crossover probability is 0.75, and the coefficient of variation is 0.031. The new drilling parameter data groups generated after encoding are M1, M2, ..., M6.

[0057] Then, the newly generated downhole parameter data are respectively related to the drilling time t to establish a relationship curve, and then the time-frequency domain feature graphs of different types of downhole parameters are obtained; the time-domain feature graphs of downhole parameters are transformed into frequency-domain feature graphs by discrete Fourier transform, and then the time-domain graphs and frequency-domain graphs of downhole parameters are fused by short-time Fourier transform to obtain the time-frequency domain feature images of different types of downhole parameters. The data enhancement function in the neural network is used to randomly flip, rotate clockwise or counterclockwise, change brightness, and increase pixel values ​​for the time-frequency domain feature graphs of each type of downhole parameter, and expand the time-frequency domain image library of downhole parameters until the time-frequency domain feature images corresponding to each type of downhole parameter are ≥1200, and the iterative calculation of the neural network is stopped.

[0058] Finally, the time-frequency domain feature maps of the drilling information of rocks with different porosities are input into the first convolution layer of the rock porosity VG-CNN prediction model, which consists of 42 convolution kernels of size 3×3. The feature maps after the first convolution are input into the first pooling layer of the VG-CNN prediction model, which consists of a 2×2 pooling window. Steps 4.1 and 4.2 are repeated until the learning rate of the VG-CNN prediction model is ≤3×10 -5 , attenuation coefficient ≤ 1×10 -6 , stop model training; at this time, the optimal hyperparameters of the model are shown in Table 1:

[0059] Table 1 Optimal hyperparameters of the rock porosity VG-CNN prediction model

[0060]

[0061]

[0062] After the VG-CNN model training is completed, a new batch of coarse-grained sandstone is selected and processed into 15cm×15cm×15cm cubic specimens. Laboratory drilling tests are carried out, and the drilling speed is set to 60mm / min, the drill bit speed is set to 200r / min, and the drilling depth is uniformly preset to 120mm. The real-time drilling parameters and time-frequency domain feature maps collected in the laboratory are input into the VG-CNN prediction model, and the porosity of the coarse-grained sandstone is obtained by inversion to be 23.15%; at the same time, the laboratory mercury injection test is used to carry out relevant tests on the true porosity of the coarse-grained sandstone. According to the test results, the porosity of the coarse-grained sandstone is 25.37%. It can be seen that the rock porosity obtained by the inversion of the present invention is relatively close to the true porosity measured in the laboratory. The present invention can accurately predict the rock porosity based on the inversion of drilling parameters.

[0063] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A rock porosity prediction method based on small sample while-drilling data inversion, characterized in that: The following steps are involved: Step S1: Collecting drilling information of the drilling rig in the laboratory or underground tunnel during drilling, and obtaining key drilling parameters such as torque M, thrust F, rotation speed N, thrust speed V, drill pipe amplitude A, and vibration acceleration a; Step S2: Perform multiple noise reduction processes on the different types of while-drilling parameters collected, and input the while-drilling parameter values ​​after multiple noise reduction into the "BP-GA" model to obtain a new while-drilling parameter derived data set; The implementation process of multiple noise reduction methods for different types of drilling parameters includes the following steps: Step 1.1: Based on the raw data of different types of drilling parameters collected continuously, calculate the extreme point m of each type of drilling parameter raw data i , average value n i and the local amplitude a i ; The local mean function a(t) is used to separate the original while-drilling data x(t), and then the envelope estimation function m(t) is used to demodulate the original while-drilling data x(t): h(t)=x(t)-a(t)(3) Repeat the above operation until the original while-drilling data x(t) is reduced to a monotonic function or has ≤3 extreme points, then stop the iterative calculation. At this time, the original while-drilling data x(t) is decomposed into k decision components PF and a decomposition residual; Step 1.2: Taking kurtosis H1, correlation coefficient H2 and root mean square value H3 as the selection criteria of PF components, the evaluation index V of PF components is ij It is expressed as: In the formula, j is 1, 2, 3, ..., k; Here we define V ij >90% of the PF components are effective components; Step 1.3: Use the wavelet denoising method to further denoise the filtered PF components, including wavelet denoising of the effective PF components of torque M, thrust F, rotation speed N, propulsion speed V, drill pipe amplitude A, and vibration acceleration a. The wavelet denoising function used for torque and rotation speed is the db3 wavelet denoising function, the wavelet denoising function used for thrust and propulsion speed is the coif3 wavelet denoising function, and the wavelet denoising function used for drill pipe amplitude and vibration acceleration is the bior3.3 wavelet denoising function. Step 1.4: Input the multiple denoised LWD parameter data into the "BP-GA" model for iteration, crossover, and mutation to achieve data enhancement and data derivation; The specific method of establishing the "BP-GA" model resume is: Step 2.1: Use the neural network BP to determine the collected while-drilling information samples and expected output values, and use the adaptive algorithm to determine the initial weight and threshold of the BP neural network corresponding to each type of while-drilling parameter, so as to eliminate unreasonable data; Step 2.2: For the six types of drilling parameters after eliminating unreasonable data, genetic algorithm is used for iteration, crossover and data mutation. The new drilling parameter data groups generated after encoding are M1, M2, ..., M6; Step 2.3: Establish relationship curves between the newly generated while-drilling parameter data and the drilling time t, and then obtain time-frequency domain characteristic diagrams of different types of while-drilling parameters; Step 2.4: Using the data enhancement function in the neural network, the time-frequency domain feature images of each type of drilling parameters are randomly flipped, rotated clockwise or counterclockwise, changed in brightness, and pixel value increased to expand the time-frequency domain image library of the drilling parameters. When the number of time-frequency domain feature images corresponding to each type of drilling parameters is ≥ 1200, the iterative calculation of the neural network is stopped; Step S3: analyzing the variation curves of different types of while-drilling parameters with the drilling time t in the new while-drilling parameter data set to obtain the time-frequency domain graph of the while-drilling parameters, and obtaining the time-frequency domain image library of different types of while-drilling parameters through the data enhancement function in the neural network; Step S4: construct a rock porosity convolutional neural network prediction model, learn and train the time-frequency domain feature image library of drilling parameters during the drilling process of rocks with different porosities, determine the initial learning rate and attenuation coefficient of the model, and obtain the rock porosity prediction result.

2. A rock porosity prediction method based on small sample while drilling data inversion as claimed in claim 1, characterized in that: In step S3, the method for obtaining time-frequency domain characteristic graphs of different types of drilling parameters includes the following steps: Step 3.1: Based on the drilling information processed by multiple noise reduction and "BP-GA" model, the variation curves of different types of drilling parameters with drilling time t are established, that is, the time domain characteristic diagram of the drilling parameters is obtained; Step 3.2: Use discrete Fourier transform to transform the time domain characteristic diagram of the drilling parameters into the frequency domain characteristic diagram. The change equation is: Where Y(n) represents a finite-length discrete frequency sequence of length L, n = 0, 1, ..., L-1, Hz, and y(t) represents a discrete time sequence of length T; Step 3.3: The time domain and frequency domain graphs of the drilling parameters are fused through short-time Fourier transform; that is, with time as the horizontal axis and the acquisition frequency and instantaneous energy change of the drilling parameters as the vertical axis, the time-frequency domain feature images of different types of drilling parameters are obtained, and the expanded time-frequency domain image library of the drilling parameters is input into the rock porosity convolutional neural network prediction model, namely the VG-CNN prediction model.

3. A rock porosity prediction method based on small sample while drilling data inversion as claimed in claim 2, characterized in that: In step S4, the specific prediction process of the rock porosity convolutional neural network prediction model is as follows: Step 4.1 First, the time-frequency domain feature maps of the drilling information of rocks with different porosities are input into the first convolutional layer of the VG-CNN prediction model, which consists of 42 convolutional kernels of size 3×3; Step 4.2 Secondly, the feature map after the first convolution is input into the first pooling layer of the VG-CNN prediction model, which consists of a 2×2 pooling window; Step 4.3 Repeat steps 4.1 and 4.2 until the learning rate of the VG-CNN prediction model is ≤ 3×10 -5 , attenuation coefficient ≤ 1×10 -6 , stop model training; at this time, substitute the rock porosity prediction confusion matrix data into the following formula to verify the prediction accuracy of the VG-CNN prediction model for rock porosity; Where: P T represents a true example, N T represents a true counterexample, P F represents a false positive, N F Indicates a false counterexample.

Citation Information

Patent Citations

  • Lithology while drilling and reservoir characteristics recognizing method

    CN101575970A

  • Porosity prediction method based on depth domain prestack angle gather

    CN106556863A

  • Unconventional tight sandstone reservoir porosity prediction method based on machine learning

    CN114397711A

  • While-drilling lithology intelligent identification method and system based on acoustic vibration signals

    CN114492521A

  • Method for determining predicted parameters of surface subsidence of mine goaf

    CN116822328A