An intelligent real-time rock inversion identification method based on while-drilling parameter characteristics

Abstract

The application discloses an intelligent real-time rock stratum inversion identification method based on a while-drilling parameter feature, which comprises the following steps: simulating drilling tests on different kinds of rock samples on an indoor drilling machine test platform, collecting while-drilling parameters in the drilling process of the rock mass by means of installing sensors on the drilling machine, performing time-frequency analysis and processing on the collected drill rod axial vibration signals and training, so as to obtain an artificial neural network model for identifying rock strata; inputting the collected while-drilling data into the trained artificial neural network model on a drilling operation site, and outputting identification results of the interface and lithology types of the drilled rock strata in real time by the artificial neural network model, so as to classify the stratum geology and rock-soil bodies, and the drilling parameters and path can be adjusted in time by the staff according to the while-drilling rock stratum identification results in the drilling process, which is favorable for guaranteeing the drilling quality, improving the drilling efficiency and saving a large amount of manpower and material resources.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geological exploration, and particularly relates to an intelligent real-time rock stratum inversion identification method based on a while-drilling parameter feature. BACKGROUND

[0002] In different engineering fields, relevant geological exploration and analysis is the most complex and arduous task. In each stage of engineering geological exploration, various geological information can be obtained from the field, including surface topography, stratum interface, fault, underground water level, weathering layer thickness distribution, and various geophysical and geochemical exploration data. However, these information are discrete and discontinuous data, and it is difficult for geological workers to directly use them to analyze the distribution law thereof in a geological body. Therefore, how to use the measured data to infer the distribution law and complex relationship of the geological information in the study area is a problem. Among them, drilling and geophysical exploration methods are the most common exploration means in engineering geological exploration. The drilling method can accurately and directly reveal abnormal geological bodies by extracting entity samples, but the extraction of entities for rock stratum identification is time-consuming and high in cost. If the type of rock stratum can be identified in real time during drilling, the drilling path and drilling parameters can be adjusted in time according to the situation, which is of great significance to ensure the quality of engineering geological exploration drilling and the smooth implementation of the project. SUMMARY

[0003] Therefore, the purpose of the present application is to provide an intelligent real-time rock stratum inversion identification method based on a while-drilling parameter feature to solve the technical problem of real-time reflection of the type of drilling rock stratum during drilling.

[0004] The intelligent real-time rock stratum inversion identification method based on a while-drilling parameter feature of the present application comprises the following steps:

[0005] 1) Simulated drilling tests are carried out on different types of rock samples on a drilling test bed, and sensors are arranged on the drilling machine to collect while-drilling parameters, including drill pipe torque, drilling pressure, drilling speed, drill pipe rotation speed and drill pipe axial vibration signals during drilling;

[0006] ① The drilling specific work is used to reflect the comprehensive characteristics of rock crushing during drilling, and the formula is as follows:

[0007] In the formula, P f is the work done by the drill bit to crush the rock per unit time, which is a function of hole depth (or stratum depth) h; D b is the outer diameter of the drill bit; D i is the inner diameter of the drill bit, D i = 0 when it is a destructive drilling; and V is the perforation rate.

[0008] 2) processing the collected drill pipe axial vibration signal:

[0009] ① the collected drill pipe axial vibration signal is processed by threshold interception method, all data with amplitude less than the set threshold are deleted, and all data greater than the set threshold are retained;

[0010] ② the retained drill pipe axial vibration signal is processed by frame, each unit sampling time period t is regarded as an observation unit, each observation unit contains N time sampling points, and the N time sampling points in an observation unit form a frame data x(n), wherein 0≤n<N;

[0011] ③ feature parameters are extracted from each frame data, the feature parameters include: variance, root mean square value, peak factor and kurtosis, and the expressions of the signal feature parameters are as follows:

[0012] Variance:

[0013] Root mean square value:

[0014] Peak factor:

[0015] Kurtosis:

[0016] In the above formula, N represents the number of time sampling points, x i represents the amplitude of the i-th sampling point in a frame of vibration data, x represents the average value of the vibration amplitude of the N sampling points in a frame of vibration data, |x i | max represents the maximum amplitude of the N sampling points in a frame of vibration data, x rms is the waveform effective value of a frame of vibration data;

[0017] The extracted variance and root mean square value represent the average energy of the vibration signal, which is used to quantitatively reflect the change of the rock interface; the extracted peak factor represents the difference degree of hard rock and soft rock, and the kurtosis index reflects the change rate of hard rock and soft rock, and the peak factor and kurtosis are used to qualitatively divide the rock properties;

[0018] ④ time domain feature analysis is performed on each frame data to obtain the logarithmic energy used to distinguish the structure characteristics of rock strata, which includes the following steps:

[0019] a) the drill pipe axial vibration signal is processed by frame, N time sampling points form an observation unit x(n), 0≤n<N, which is called a frame; adjacent two frames overlap W points, and the value of W is

[0020] b) fast Fourier transform is performed on each frame data to calculate the frequency spectrum:

[0021]

[0022] c) Calculate the log energy of each frame of data:

[0023] d) Calculate the log energy after passing the amplitude spectrum of the log energy through a set of filters, the filters being triangular filters, select a set of center frequencies f m , m = 1, 2,..., M, the triangular filter frequency response being:

[0024] where,

[0025] the filtered log energy

[0026] 3) The same frame processing of the drilling specific work and the drill pipe axial vibration signal obtained by collecting the while-drilling parameters is carried out, and the time domain characteristic indexes of the vibration signal are analyzed through the variance, the root mean square value, the crest factor and the kurtosis;

[0027] 4) The variance, the root mean square value, the crest factor and the kurtosis of the drilling specific work and the drill pipe axial vibration signal in the same time unit, the log energy E and the filtered log energy S(m) of the drill pipe axial vibration signal in the same time unit are taken as an array, and all the data are used to form a training set for training the artificial neural network;

[0028] 5) An artificial neural network model is constructed, the artificial neural network model being composed of an input layer, a hidden layer and an output layer, the artificial neural network model being trained by using the data in the training set in step 4), the training algorithm being the KFCM algorithm, and an artificial neural network model for identifying the rock layer category encountered during drilling being obtained; the clustering objective function of the KFCM algorithm is:

[0029]

[0030]

[0031] In the formula, k is the number of categories; N is the number of samples in the original characteristic space; x j is the jth sample in the original characteristic space, j = 1, 2,..., N; μ ij is the jth sample x j The membership μ ij of the ith category ∈ [0, 1]; m is the fuzziness; v i is the clustering center of the ith category in the high-dimensional characteristic space, i = 1, 2,..., k; d(φ(x j ), v i ) is the distance between the jth sample xj distance between the cluster center v of the i-th type and the cluster center v of the j-th type; then the cluster center v of the KFCM algorithm is i i

[0032]

[0033] The minimum objective function can be derived by setting the partial derivative of it to the membership matrix U to 0, then the membership μ is rs

[0034]

[0035] In the formula, r=1, 2,..., N; s=1, 2,..., k;

[0036] The cluster center v and the membership μ can be continuously optimized and calculated through formula (3) and (4), so as to obtain the clustering result. i rs

[0037] 6) The fuzzy C-means clustering algorithm is used for stratigraphic interface classification and identification. The algorithm is defined as follows: assuming that X={x1, x2,..., x N} is a drilling parameter drilling specific work and a collected drill pipe characteristic parameter data set in a drilling process to be clustered, the FCM algorithm assigns each object x j (1≤j≤N) to C stratigraphic classes, and uses a membership matrix U={u ij} of C*N to represent the membership relationship between objects and classes, wherein u ij indicates the membership degree of the j-th object to the i-th class. The fuzzy partition space defined for X is:

[0038]

[0039] 7) The sensors are installed on the drilling rig on the drilling site to collect the drilling rig torque, feed pressure, drilling speed, drilling rig rotating speed and other parameters in real time, the real-time drilling specific work is obtained through calculation, the collected drill pipe axial vibration signal data are frame processed, then the characteristic parameters of each frame data, such as variance, root mean square value, wave peak factor and kurtosis, and the logarithmic energy E and the filtered logarithmic energy S(m) of the axial vibration signal in each frame data are extracted, the data are input into the trained artificial neural network model in real time, and the artificial neural network model outputs the clustering analysis result of the real-time drilled rock formation.

[0040] The beneficial effects of the present application are:

[0041] ​​​​​The application is based on an intelligent real-time rock stratum inversion identification method based on drilling parameter characteristics. Different types of rocks are drilled for drilling tests, and corresponding drilling parameters are collected, especially after a series of processing of the collected drill pipe vibration data, vibration signal characteristics reflecting rock characteristics are obtained, and a data set for training an artificial neural network model is built by using these signal characteristics. The drilling parameters collected in real time during the actual drilling process are processed by the trained artificial neural network model, the type of the drilled rock stratum is identified, the rock stratum conditions at different depths of the drill hole are accurately obtained in real time, the stratum geology and rock-soil body classification are performed, and the drilling parameters and path can be adjusted in time according to the drilling rock stratum identification result during drilling by the staff, which is beneficial to guiding the drilling operation during drilling, adjusting the drilling parameters and path in time according to the rock stratum type, ensuring the drilling quality, improving the drilling efficiency, saving a large amount of manpower and material resources, and also being an important direction of the intelligent development of drilling engineering. DETAILED DESCRIPTION

[0042] The application will be further described below in combination with examples.

[0043] The intelligent real-time rock stratum inversion identification method based on drilling parameter characteristics in the embodiment has the following characteristics:

[0044] 1) Simulated drilling tests are carried out on different types of rock samples on a drilling test bench, and sensors are arranged on a drilling machine to collect drilling parameters, the drilling parameters including drill pipe torque, drilling pressure, drilling speed, drill pipe rotating speed and drill pipe axial vibration signal during drilling;

[0045] ①Drilling specific work is used to reflect the comprehensive characteristics of rock crushing during drilling, and the formula is as follows:

[0046] In the formula, P f is the work done by the drill bit to crush rocks per unit time during drilling, which is a function of hole depth (or stratum depth) h; D b is the outer diameter of the drill bit; D i is the inner diameter of the drill bit, when it is a destructive drilling, D i = 0; V is the perforation rate.

[0047] 2) The collected drill pipe axial vibration signal is processed:

[0048] ①The collected drill pipe axial vibration signal is processed by threshold interception method, all data with an amplitude less than a set threshold are deleted, and all data greater than the set threshold are retained;

[0049] ②Frame processing is performed on the retained axial vibration signal of the drill pipe, each unit sampling time period t is regarded as an observation unit, N time sampling points are contained in each observation unit, the N time sampling points in one observation unit form a frame data x(n), wherein 0≤n<N;

[0050] ③Feature parameters are extracted from each frame data, the feature parameters include: variance, root mean square value, wave crest factor and kurtosis, the expressions of the signal feature parameters are as follows:

[0051] Variance:

[0052] Root mean square value:

[0053] Wave crest factor:

[0054] Kurtosis:

[0055] In the above formulae, N represents the number of time sampling points, x i represents the amplitude of the i-th sampling point in a frame of vibration data, represents the average value of the vibration amplitudes of the N sampling points in a frame of vibration data, |x i | max represents the maximum amplitude in the N sampling points in a frame of vibration data, x rms represents the waveform effective value of a frame of vibration data;

[0056] The extracted variance and root mean square value represent the average energy of the vibration signal, and are used for quantitatively reflecting the change of the rock interface; the extracted wave crest factor represents the difference degree of hard rock and soft rock, and the kurtosis index reflects the change rate of hard rock and soft rock, and the wave crest factor and the kurtosis are used for qualitatively dividing the rock properties;

[0057] ④Time domain feature analysis is performed on each frame data to obtain the logarithmic energy used for distinguishing the structure characteristics of the rock stratum, which includes the following steps:

[0058] a) Frame processing is performed on the axial vibration signal of the drill pipe, N time sampling points form an observation unit x(n), 0≤n<N, which is called a frame; adjacent two frames overlap W sampling points, and the value of W is

[0059] b) Fast Fourier transform is performed on each frame data to calculate the frequency spectrum:

[0060]

[0061] c) The logarithmic energy of each frame data is calculated:

[0062] d) passing the amplitude spectrum of the log energy through a set of filters, calculating the log energy after passing through the filters, the filters being triangular filters, selecting a set of center frequencies f m , m = 1, 2,..., M, the triangular filter frequency response being:

[0063] wherein,

[0064] the filtered log energy

[0065] 3) the same frame processing of the drilling specific work and the drill pipe axial vibration signal obtained by collecting the while-drilling parameters is performed, and the time domain characteristic indexes of the vibration signal are analyzed through variance, root mean square value, wave crest factor and kurtosis;

[0066] 4) the variance, root mean square value, wave crest factor and kurtosis of the drilling specific work and the drill pipe axial vibration signal in the same time unit, and the log energy E and the filtered log energy S(m) of the drill pipe axial vibration signal in the same time unit are taken as an array, and all the data are used to form a training set for training the artificial neural network;

[0067] 5) an artificial neural network model is constructed, the artificial neural network model being composed of an input layer, a hidden layer and an output layer, the artificial neural network model being trained by using the data in the training set in step 4), the training algorithm being the KFCM algorithm, and an artificial neural network model for identifying the rock layer category encountered during drilling is obtained; the clustering objective function of the KFCM algorithm is:

[0068]

[0069]

[0070] wherein, k is the number of categories; N is the number of samples in the original feature space; x j is the jth sample in the original feature space, j = 1, 2,..., N; μ ij is the jth sample x j of the ith category; μ ij ∈ [0, 1]; m is the fuzziness; v i is the clustering center of the ith category in the high-dimensional feature space, i = 1, 2,..., k; d(φ(x j ), v i ) is the distance between the jth sample x j and the clustering center v i of the ith category in the high-dimensional feature space; and the clustering center v i of the KFCM algorithm is

[0071]

[0072] The objective function can be minimized by taking its derivative with respect to the membership matrix U as 0, then the membership μ rs for:

[0073]

[0074] Where r = 1, 2, ..., N; s = 1, 2, ..., k;

[0075] The cluster center v can be calculated by continuously optimizing through equations (3) and (4) i and membership μ rs , thus obtaining the clustering results;

[0076] 6) Fuzzy C-means clustering algorithm is used to classify and identify the stratum interface. The algorithm is defined as follows: Assume that X = {x1, x2, ..., x N} is the drilling parameters and drilling power ratio in the drilling process to be clustered, and the drill pipe characteristic parameter data set collected. The FCM algorithm converts each object x j (1≤j≤N) are assigned to C stratum categories, using the C×N membership matrix U={u ij} represents the affiliation between the object and the category, where u ij Indicates the degree of membership of the j-th object to the i-th class. The fuzzy partition space for X is defined as:

[0077]

[0078] 7) Sensors are installed on the drill rig at the drilling site to collect parameters such as drill pipe torque, feed pressure, drilling speed, and drill rig speed in real time. Real-time drilling power ratio is calculated and the collected drill pipe axial vibration signal data is framed. Feature parameters are then extracted from each frame of data: variance, root mean square value, crest factor, and kurtosis. The logarithmic energy E and filtered logarithmic energy S(m) of the axial vibration signal in each frame of data are also extracted. The data are input into a trained artificial neural network model in real time, and the artificial neural network model outputs cluster analysis results of the rock formations encountered in real time.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for intelligent real-time lithology inversion identification based on drilling parameter characteristics, characterized in that: The method comprises the following steps: 1) performing simulation drilling tests on different kinds of rock samples on a drilling test platform, and installing sensors on a drilling rig to collect drilling parameters, including drill pipe torque, drilling pressure, drilling speed, drill pipe rotation speed and drill pipe axial vibration signals during drilling; ① using drilling specific work to reflect the comprehensive characteristics of rock breaking during drilling, obtaining the change of drilling specific work with drill bit displacement in the drilled stratum, and the formula is: where P f is the work done by the bit to break rock per unit time during drilling; D b is the bit outside diameter; D i is the bit inside diameter, D i = 0 when drilling is destructive; and V is the rate of penetration. 2) processing the collected drill pipe axial vibration signals: ① using threshold interception method to process the collected drill pipe axial vibration signals, deleting all data with an amplitude less than a set threshold, and retaining all data greater than the set threshold; ② performing frame processing on the retained drill pipe axial vibration signals, regarding each unit sampling time period t as an observation unit, and each observation unit containing N time sampling points, and the N time sampling points in an observation unit forming a frame of data x(n), wherein 0≤n<N; ③ extracting feature parameters from each frame of data, and the feature parameters include: variance, root mean square value, wave crest factor and kurtosis, and the expressions of the signal feature parameters are as follows: variance: Root mean square value: Peak factor: kurtosis: N represents the number of time sampling points, x i represents the amplitude of the i-th sampling point in a frame of vibration data, represents the average value of the amplitudes of the N sampling points in a frame of vibration data, |x i | max represents the maximum amplitude of the N sampling points in a frame of vibration data, x rms represents the waveform effective value of a frame of vibration data; The extracted variance and root mean square value represent the average energy of the vibration signal, and are used to quantitatively reflect the change of the rock interface; the extracted wave crest factor represents the difference degree of hard rock and soft rock, and the kurtosis index reflects the change rate of hard rock and soft rock, and the wave crest factor and kurtosis are used to qualitatively divide the rock properties; ④ performing time domain feature analysis on each frame of data to obtain the logarithmic energy used to distinguish the structural characteristics of the rock mass, which comprises the following steps: a) Frame the axial vibration signal of the drill pipe, N time sampling points form an observation unit x(n), 0≤n<N, called a frame; the adjacent two frames overlap W points, and the value of W is b) performing fast Fourier transform on each frame of data to calculate the frequency spectrum: c) calculating the log energy of each frame of data: d) passing the amplitude spectrum of the log energy through a set of filters, calculating the log energy after passing through the filters, the filters being triangular filters, selecting a set of center frequencies f m , m = 1, 2,..., M, the triangular filter frequency response being: filtered log energy 3) performing frame processing on the drilling specific work obtained by collecting the drilling parameters and the drill pipe axial vibration signals, and analyzing the time domain feature indexes of the vibration signals through variance, root mean square value, wave crest factor and kurtosis; 4) taking the variance, root mean square value, wave crest factor and kurtosis of the drilling specific work and the drill pipe axial vibration signals in the same time unit, and the logarithmic energy E and the filtered logarithmic energy S(m) of the drill pipe axial vibration signals in the same time unit as an array, and taking all the data to form a training set for training the artificial neural network; 5) constructing an artificial neural network model, which comprises an input layer, a hidden layer and an output layer, training the artificial neural network model using the data in the training set in step 4), and using KFCM algorithm as the training algorithm to obtain an artificial neural network model for identifying the category of drilled stratum; the clustering objective function of the KFCM algorithm is: Where k is the number of cluster categories; N is the number of samples in the original feature space; x j is the jth sample in the original feature space, j = 1, 2..., N; μ ij is the jth sample x j The membership degree μ to the i-th class ij ∈[0, 1]; m is the fuzziness; v i is the cluster center of the i-th class in the high-dimensional feature space, i=1,2...,k; d(φ(x j ), v i ) is the jth sample x in the high-dimensional feature space j and the cluster center v of the i-th class i The distance between them; then the cluster center v of the KFCM algorithm i for The minimization of the objective function can be performed by taking the partial derivative of it with respect to the membership matrix U, setting it to zero, and solving for the membership μ rs : In the formula, r=1, 2,..., N; s=1, 2,..., k; The cluster centers v are calculated by continuously cycling optimization of formulas (3) and (4) i And membership μ rs So as to obtain the clustering result; 6) Fuzzy C-means clustering algorithm is used to classify and identify the stratum interface; the algorithm is defined as follows: Assume that X = {x1, x2, ..., x N } is the drilling parameter drilling power and the drill pipe characteristic parameter data set collected during the drilling process to be clustered. The FCM algorithm converts each object x j (1≤j≤N) are assigned to C stratum categories, using the C×N membership matrix U={u ij } represents the affiliation between the object and the category, where u ij Indicates the degree of membership of the j-th object to the i-th class; the fuzzy partition space for X is defined as: 7) On the drilling rig at the drilling site, sensors are installed to collect real-time drilling pipe torque, feed pressure, drilling speed and drilling rig speed to obtain real-time drilling specific work through calculation, and the collected drilling pipe axial vibration signal data is processed by frames, then the characteristic parameters of each frame data are extracted: variance, root mean square value, wave peak factor and kurtosis, and the logarithmic energy E and the filtered logarithmic energy S(m) of the axial vibration signal in each frame data are extracted, the data is input into the trained artificial neural network model in real time, and the artificial neural network model outputs the clustering analysis result of the real-time drilled rock formation.

Images