Lithology key indicative feature extraction method for vibration high-frequency signal during drilling

Abstract

The invention belongs to the field of intelligent rock drilling and digging, and particularly relates to a method for extracting lithology key indicative features of vibration high-frequency signals during drilling. The feature extraction method provided by the invention comprises the following steps: step 1, collecting a while-drilling process vibration high-frequency signal generated in a drilling process in real time; 2, preprocessing the vibration high-frequency signal in the process while drilling; and step 3, extracting time domain, frequency domain and time frequency information from the preprocessed vibration signal. The relevance between the time domain information, the frequency domain information and the time frequency information extracted through the method and the lithology is higher, and lithology classification can be better achieved. Therefore, the method has a good application prospect.

Description

Technical Field

[0001] The invention belongs to the field of intelligent rock drilling, and in particular relates to a method for extracting key indicative features of lithology from high-frequency vibration signals during a drilling process. Background Art

[0002] Rock drilling is an important means of underground space development. During the drilling process, identifying the rock properties is an important engineering technical problem. Rock property identification is of great significance in optimizing drilling design and construction, preventing accidents, geological research, environmental protection and subsequent operation decision-making.

[0003] Traditional lithology identification methods require coring of rocks during drilling, which is a complex process and affects the overall efficiency of the project. Therefore, if a technology that can automatically identify lithology during drilling can be developed, it can effectively ensure drilling efficiency and coring quality, and reduce the number of corings, which is of great significance to the development of the entire geological survey technology. Deep drilling vibration signals are changes in dynamic mechanical motion parameters caused by factors such as geological heterogeneity, drill tool interaction, and mechanical operation imbalance during drilling. These fluctuations include drill string speed, acceleration, impact, swing load, and noise, which can reflect formation characteristics and drilling conditions. Therefore, deep drilling vibration signals have the potential to be used to extract features and perform lithology identification.

[0004] The interaction between the drill bit and the rock is a complex mechanical process, and vibration is the physical information stimulated in this process, so the vibration signal is a carrier of comprehensive information. When collecting vibration signals, if we only consider the differences caused by different lithologies and ignore the influence of factors such as drilling parameters, it will inevitably limit the generalization ability of the lithology intelligent identification model.

[0005] In the actual drilling process, the drill pipe speed affects the cutting torque and cutting rate of the drill bit, and the diamond impregnated on the drill bit will also affect its wear. In addition, the drill pipe speed will also affect the stability of the drill pipe itself; in actual drilling, the mechanical drilling speed is determined by the formation characteristics, drilling pressure and drilling fluid performance. All of the above factors directly affect the mechanical action process between the drill bit and the rock; at the same time, the drill bit diameter also directly affects the contact area and crushing volume between the drill bit and the rock, and will also affect the wear of the diamond. Therefore, the impact of these factors will eventually act on the vibration signal of the drilled rock. The process of establishing the correlation between the formation generalization model and the vibration signal is particularly cumbersome and complicated. It can be seen that how to extract features with strong correlation with lithology from the deep drilling vibration signal is the key issue to achieve a more accurate and generalized lithology intelligent identification model. Summary of the invention

[0006] Based on the above problems, the present invention provides a method for extracting key indicative features of lithology from high-frequency vibration signals during drilling.

[0007] A method for extracting key indicative features of lithology from high-frequency vibration signals during drilling, comprising the following steps:

[0008] Step 1, real-time acquisition of high-frequency vibration signals generated during drilling;

[0009] Step 2, preprocessing the high-frequency vibration signal during the drilling process;

[0010] Step 3: extract the time domain, frequency domain and time-frequency information from the preprocessed vibration signal.

[0011] Preferably, in step 1, information on the corresponding rotation speed, drilling speed and drill bit diameter when the high-frequency vibration signal during drilling is generated is also collected.

[0012] Preferably, the rotation speed ranges from 200 to 400 RPM; the drilling speed ranges from 1 to 3 mm / min; the drill bit diameter ranges from 25 to 40 mm; and the drilling depth ranges from 0 to 20 mm.

[0013] Preferably, in step 2, the specific method of preprocessing is selected from at least one of the following methods: pre-emphasis, wavelet threshold denoising and bandpass filtering.

[0014] Preferably, in step 2, the wavelet threshold denoising selects the Sym8 wavelet basis function and performs 9-layer wavelet decomposition on the signal.

[0015] Preferably, in step 3, before extracting the time domain, frequency domain and time-frequency information, it is verified that the vibration signal is a wide stationary random signal.

[0016] Preferably, in step 3, the method for verifying that the vibration signal is a wide stationary random signal comprises the following steps:

[0017] Step a: Divide samples of different observation scales into N subsamples with specific frame lengths and frame shifts, and calculate the mean μ of each subsample i ;

[0018] Step b, calculate the sample mean Record the successive μ in the subsample i <μ or μ i >μ is a run, and the total number of runs is R; i <μThe number of subsamples is m, satisfying μ i >The number of subsamples of μ is n;

[0019] Step c, assume that the subsample mean μ iis a constant, and the acceptance region is tested using the normal distribution. The test statistic Z is calculated according to the following formula:

[0020]

[0021] where R is the numerator part in the test statistic, and μ R is the sample mean, σ R is the sample standard deviation, and the meaning of ~N(0,1) is the normal distribution;

[0022]

[0023] where m is the number of subsamples that satisfy u i < u, and n is the number of subsamples that satisfy u i > u;

[0024] Step d, set a two-sided test at a significance level of α = 0.05. The acceptance region of the null hypothesis is |Z| < Z α / 2 , and when |Z| < 1.96, the null hypothesis is accepted. The subsample μ i is a constant, that is, the original random signal is a wide-sense stationary random signal.

[0025] Preferably, in step 3, the time-domain information is extracted by short-time Fourier transform.

[0026] The present invention also provides a system for implementing the above feature extraction method, which is characterized by including:

[0027] A signal acquisition module for real-time acquisition of high-frequency vibration signals during the drilling process;

[0028] A preprocessing module for preprocessing the high-frequency vibration signals during the drilling process;

[0029] A feature extraction module for extracting time-domain, frequency-domain, and time-frequency information from the preprocessed vibration signals.

[0030] The present invention also provides a computer-readable storage medium, on which there is stored: a computer program for implementing the above feature extraction method.

[0031] After adopting the technical solution of the present invention, the following beneficial effects can be achieved:

[0032] 1. The present invention collects the vibration signals of seven types of rocks in the drilling process in real time under a comprehensive test based on three factors and three levels for data processing. The signals are pre-emphasized and denoised by wavelet threshold and then analyzed in time domain and frequency domain to obtain the data set required for building the lithology recognition model. After time domain, frequency domain and time-frequency analysis, a method for extracting deep while-drilling vibration signal features is proposed. The feature extraction method of the present invention can effectively improve the accuracy and reliability of while-drilling vibration signal acquisition during drilling.

[0033] 2. The present invention fully considers the differences in vibration signals caused by rock mineral effects, mechanical effects and changes in drilling parameters during the mutual feedback between rock and drill bit, and provides a feature extraction method that can effectively extract features related to lithology.

[0034] Obviously, according to the above contents of the present invention, in accordance with common technical knowledge and customary means in the art, without departing from the above basic technical ideas of the present invention, other various forms of modification, replacement or change may be made.

[0035] The above contents of the present invention are further described in detail below through specific implementation methods in the form of embodiments. However, this should not be understood as the scope of the above subject matter of the present invention being limited to the following examples. All technologies realized based on the above contents of the present invention belong to the scope of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 The signal time domain characteristics of granite at different rotation speeds;

[0037] Figure 2 The signal time domain characteristics of granite at different drilling speeds;

[0038] Figure 3 The signal time domain characteristics of granite with different drill bit diameters;

[0039] Figure 4 The signal time domain characteristics at different drilling depths of granite;

[0040] Figure 5 The time domain characteristics of vibration signals of seven types of lithology;

[0041] Figure 6 is the power spectrum density at different rotation speeds and drilling speeds;

[0042] Figure 7 is the power spectrum density for different drill diameters and drilling depths. DETAILED DESCRIPTION

[0043] It should be noted that the algorithms of data collection, transmission, storage and processing steps not specifically described in the embodiments, as well as the hardware structure, circuit connection, etc. not specifically described can all be implemented through the contents disclosed in the prior art.

[0044] Embodiment 1 Method and system for extracting key indicative features of lithology from high-frequency vibration signals during drilling The feature extraction system of this embodiment includes:

[0045] A signal acquisition module, used for real-time acquisition of high-frequency vibration signals generated during drilling;

[0046] A preprocessing module, used for preprocessing the vibration high-frequency signal during the drilling process;

[0047] The feature extraction module is used to extract the time domain, frequency domain and time-frequency information from the preprocessed vibration signal.

[0048] The method for extracting key indicative features of lithology of high-frequency vibration signals during drilling by the above system comprises the following steps:

[0049] Step 1, real-time acquisition of the high-frequency vibration signal of the drilling process generated during the drilling process. In addition, the corresponding rotation speed, drilling speed and drill bit diameter information when the high-frequency vibration signal of the drilling process is generated are recorded simultaneously during the information acquisition. As a preferred solution, the rotation speed ranges from 200 to 400 RPM; the drilling speed ranges from 1 to 3 mm / min; the drill bit diameter ranges from 25 to 40 mm; and the drilling depth ranges from 0 to 20 mm.

[0050] Step 2, preprocessing the high-frequency vibration signal during the drilling process; the specific method of the preprocessing includes: pre-emphasis, wavelet threshold noise reduction and bandpass filtering.

[0051] Step 3: extract the time domain, frequency domain and time-frequency information from the preprocessed vibration signal.

[0052] The three preprocessing methods are described as follows:

[0053] When the triaxial acceleration sensor is working, it converts the received physical signal into an analog signal represented by current or voltage. After A / D processing, the analog signal is converted into a digital signal representing vibration. However, the vibration signal at this time is not the original vibration signal. It often contains other signals, such as the environment, circuit defects, and signals generated during signal transmission. These signals will affect the reliability and accuracy of vibration signal processing and analysis. Therefore, the vibration signal needs to be preprocessed before processing and analysis to improve the signal-to-noise ratio. Based on the characteristics of the vibration signal, the filtering and noise reduction in this article includes pre-emphasis, wavelet threshold noise reduction, and bandpass filtering.

[0054] 1. Pre-emphasis

[0055] When the vibration signal propagates in the air and rocks, its energy will be lost by the dielectric, which is particularly serious in the high-frequency band of the signal. At the same time, the signal will also produce energy loss during the entire acquisition and transmission process, and the energy in the high-frequency band will be further reduced. In order to compensate for the high-frequency energy of the signal during the propagation and acquisition process, the pre-emphasis method is required to process the signal.

[0056] Pre-emphasis has no effect on noise, so compensating high-frequency components can effectively improve the signal-to-noise ratio. Pre-emphasis can filter the vibration signal with a first-order FIR high-pass filter. The transfer function of the filter in the z domain is as follows:

[0057] H(z)=1-az -1

[0058] In the formula, a is the pre-emphasis coefficient, H(z) is the transfer coefficient in the z domain, and z is the specific amplitude in this direction. a is usually in the range of 0.9<a<1. In order to facilitate the subsequent analysis of the characteristic information of the vibration signal, a=0.95.

[0059] 2. Wavelet threshold denoising

[0060] Wavelet threshold denoising is widely used in signal denoising. When noisy signals are transformed to different frequency bands through wavelet transform, the wavelet coefficients of effective signals and noise are distributed differently. The effective signal corresponds to a larger wavelet coefficient in the wavelet domain, while the remaining wavelet coefficients reflect the noise. Therefore, setting a threshold to retain a larger wavelet coefficient and filtering the remaining wavelet coefficients, and finally reconstructing the wavelet coefficients can achieve signal denoising.

[0061] Before performing wavelet threshold denoising, it is necessary to select appropriate wavelet basis functions, decomposition layers, thresholds, and threshold functions. Symlets wavelets have compact support, orthogonality, and approximate symmetry [based on the denoising study of the sym8 wavelet north-seeking system], and have good accuracy in signal processing, so the Sym8 wavelet basis function is selected for filtering and denoising of vibration signals; since the sampling frequency of the vibration signal is 51200Hz, in order to make the frequency band have a higher resolution after the wavelet transform, the signal is decomposed into 9 layers of wavelets, and the frequency resolution is 100Hz at this time; in terms of threshold quantization methods, in view of the fact that the reconstructed signal after the hard threshold function processing is prone to oscillation, and the reconstructed signal after the soft threshold function processing is deviated from the real signal, this paper uses a soft-hard compromise threshold function to process the wavelet coefficients, and the formula is as follows:

[0062]

[0063] In the formula, w j is the wavelet coefficient of the jth layer; w jnis the wavelet coefficient after the j-th layer threshold function processing; α∈[0,1], take α=0.5; T is the threshold, the calculation formula is as follows:

[0064]

[0065] Where w1 is the wavelet coefficient of the first layer; N is the length of the signal.

[0066] 3. Bandpass filtering

[0067] The response frequency range of the triaxial accelerometer is 10 to 20,000 Hz. Within the response frequency range, the voltage output by the sensor is proportional to the acceleration. At this time, the frequency response curve tends to be flat, and the measured signal can better restore the real vibration signal. To avoid signal aliasing, the Nyquist sampling theorem requires that the sampling frequency should be greater than twice the highest frequency, but considering the response frequency range of the sensor, the signal between the maximum response frequency and 1 / 2 of the sampling frequency must be filtered in the end.

[0068] When collecting vibration signals during drilling and rock crushing, it is inevitable that the mechanical vibration of the test bench will be collected at the same time. The vibration frequency of mechanical equipment is mainly distributed in the range of 10 to 5000 Hz, and its energy is concentrated in the frequency range of less than 2000 Hz [Reference: MEMS accelerometers for engineering vibration environment monitoring needs and their applications], while the vibration frequency of drill bit-rock interaction is distributed above 3000 Hz [Reference: Research on vibration signal feature recognition when the drill bit drills into different media].

[0069] Based on the above situation, a bandpass filter with a lower limit frequency of 3000Hz and an upper limit frequency of 10000Hz needs to be designed to process the vibration signal.

[0070] The Butterworth digital filter is an IIR filter with a flat frequency response curve in the passband, and has advantages such as linear phase, attenuation slope and loading characteristics. There are many types of filters such as low-pass, high-pass, band-stop, and band-pass [Reference: Online monitoring of tool wear status based on multi-source signal feature fusion]. Therefore, the vibration signal is filtered with a sixth-order Butterworth bandpass filter, and its transfer function in the z domain is as follows:

[0071]

[0072] Where a(m) and b(n) are coefficient vectors; N is the order of the bandpass filter.

[0073] Before extracting the information in the time domain, frequency domain and time-frequency domain, verify that the vibration signal is a wide-stationary random signal. The method for verifying that the vibration signal is a wide-stationary random signal is:

[0074] If the random signal meets the following conditions:

[0075] E[x(t)]=μ

[0076] E[x(t)x(t+τ)]=R(τ)

[0077] Among them, x() is the vibration signal, t is time, τ is the time difference, and E[] represents the expected value, which is usually used to represent the average value or expectation of a random variable. In vibration signal analysis, E[x(t)] can represent the average value of the vibration signal x(t) at all possible times t, reflecting the overall characteristics of the vibration signal. R() usually represents a correlation function or an autocorrelation function. In vibration signal analysis, R(t1, t2) can represent the correlation or autocorrelation of vibration signals at different time points t1 and t2. The autocorrelation function R(t) measures the autocorrelation of the signal within time t, that is, the degree of similarity between the signal and itself at different time delays; μ is the mean of the random process;

[0078] That is, if the signal mean is a constant that is independent of time and the autocorrelation function is only related to the time difference, the random signal can be called a wide-stationary random signal.

[0079] The vibration signal is judged as a wide and stable random signal, and then the stability of the vibration signal at different observation scales is analyzed, so as to provide a basis for selecting the observation scale for analyzing the signal characteristics.

[0080] The vibration signals x(t) of seven types of lithology are used as the objects to test the stability of random signals. The runs test is used for analysis. This method belongs to the non-parametric hypothesis testing method, and its steps are as follows:

[0081] Step a: divide the samples with observation scales of 1s, 2s, 4s, and 8s into N subsamples with a frame length of 30ms and a frame shift of 10ms, and calculate the mean μ of each subsample. i ;

[0082] Step b, calculate the sample mean Record the successive μ in the subsample i <μ or μ i >μ is a run, and the total number of runs is R; i <μThe number of subsamples is m, satisfying μ i >The number of subsamples of μ is n;

[0083] Step c, assume that the subsample mean μ i is a constant, and the normal distribution is used to test the acceptance domain. The test statistic Z is calculated according to the following formula:

[0084]

[0085] Among them, R is the numerator of the test statistic, which usually represents the difference between the subsample means, and μR is the sample mean, and σ R is the sample standard deviation. The meaning of ~N(0,1) is the normal distribution;

[0086]

[0087] where m is the number of subsamples satisfying u i < u, and n is the number of subsamples satisfying u i > u;

[0088] Step d, set a two-sided test at a significance level of α = 0.05. The acceptance region of the null hypothesis is |Z| < Z α / 2 By referring to the standard normal distribution table, the critical value is known to be 1.96. When |Z| < 1.96, the null hypothesis is accepted, and the subsample μ i is a constant, that is, the original random signal is a wide-sense stationary random signal.

[0089] In Step 3, for time-frequency analysis methods, currently, the short-time Fourier transform (STFT), wavelet transform (WT), and Hilbert-Huang transform (HHT) are commonly used to analyze the time-frequency domain of signals. As the earliest proposed time-frequency analysis method, the short-time Fourier transform has the advantages of simple algorithm and fast processing speed, but its time resolution and frequency resolution cannot be achieved simultaneously. Compared with the short-time Fourier transform, the wavelet transform improves the time resolution and frequency resolution, but the selection of wavelet bases is completely based on experience, without self-adaptability, and its analysis efficiency is lower than that of the short-time Fourier transform. Although the Hilbert-Huang transform solves the problem of self-adaptability, its analysis efficiency is also low, and its two types of analysis bases, empirical mode decomposition (EMD) and variational mode decomposition (VMD), respectively have problems of mode mixing and difficulty in determining coefficients. In the lithology identification task, the analysis efficiency of vibration signals is required to be relatively high. Therefore, the wavelet transform and Hilbert-Huang transform with low analysis efficiency and poor self-adaptability do not meet the conditions during time-frequency analysis. Therefore, the short-time Fourier transform is selected for time-frequency analysis.

[0090] During the time-frequency analysis using the short-time Fourier transform, to reduce spectral leakage and Gibbs phenomenon, a Hamming window w(n) with a window length of nfft is selected to intercept the original signal x(n), and the Fourier transform is performed on the windowed signal. Then, each time the window is moved by the length of hop size to make two adjacent windows overlap by the length of overlap, and the Fourier transform is continued after windowing. Finally, a two-dimensional matrix S(ω, τ) is obtained, and each column vector represents the spectrum under the corresponding window. The calculation formula is as follows:

[0091]

[0092] Where x(n) is the original signal, τ is the time offset in STFT, and ω is the frequency variable in STFT.

[0093] Since the vibration signal is a power signal, its energy should be measured by the power spectrum density. Therefore, the amplitude S(ω,τ) at the corresponding time and frequency should be converted into the power spectrum density P(ω,τ). The calculation formula is as follows:

[0094]

[0095] Where f r is the frequency resolution, f s is the sampling rate, and NFFT is the Hamming window length.

[0096] The features extracted by this method (information in the time domain, frequency domain and time-frequency) can be used to classify lithology. The seven types of rocks classified are marble, marl, limestone, quartz sandstone, granite, feldspar sandstone and shale.

[0097] The technical solution of the present invention is further illustrated by experiments below.

[0098] Experimental Example 1 Study on the correlation between time domain, frequency domain and time-frequency information and lithology

[0099] 1. Purpose of the experiment

[0100] The purpose of this experimental example is to examine the correlation between the characteristic information extracted by the method of Example 1 and the lithology, so as to verify whether the characteristic information extracted by the method of the present invention can accurately achieve the purpose of lithology classification.

[0101] 2. Experimental Methods

[0102] In this experimental example, the method for extracting time domain, frequency domain and time-frequency information is performed according to the method of Example 1. Part of the specific parameters and steps are described as follows:

[0103] The hardware used in this experiment includes:

[0104] Model ZK5040BS / 2 vertical CNC drilling machine: used as a micro-drilling test bench;

[0105] Three-axis acceleration sensor: used to fully collect rock-breaking vibration signals;

[0106] Host computer: saves the digital signal converted from analog signal by A / D conversion of data acquisition card;

[0107] Data acquisition card: Connects to the sensor, has four channels, and is used to receive analog signals from four sensor cables. Based on the Nyquist sampling theorem, the sampling frequency of the data acquisition card is set to 51200Hz.

[0108] The specific process is:

[0109] S1, in a comprehensive test of three factors and three levels (the three factors are rotation speed, drilling speed and drill bit diameter, and three levels are set for each factor), real-time acquisition of vibration signals of seven types of rocks during drilling is performed for data processing;

[0110] S2, extracting time domain, frequency domain and time-frequency information according to the method of Example 1;

[0111] S3, analyze the relationship between time domain, frequency domain and time-frequency information and lithology classification.

[0112] The seven types of rock samples are described as follows:

[0113] Table 1 Basic information of various lithologies

[0114]

[0115] According to different genesis, the seven types of lithology include marl, limestone, quartz sandstone, feldspar sandstone and shale, which are sedimentary rocks, and granite and marble, which are igneous rocks and metamorphic rocks respectively. Among them, some lithologies have the same mineral types, while some lithologies have significantly different mineral compositions. In addition to the main minerals, the rock mechanical parameters of lithologies are also different. Marl, limestone and shale have the greatest compressive strength and tensile strength, and the values ​​of the three are relatively close; the compressive strength and tensile strength of marble and granite are in the middle, and the values ​​of the two are very close; while the compressive strength and tensile strength of quartz sandstone and feldspar sandstone are significantly lower. Therefore, these seven types of lithology have similarities and differences, which is of certain significance for studying the intelligent identification method of lithology.

[0116] In order to enable the micro-drill test bench to firmly clamp the sample and reduce relative vibration, the size of each type of rock sample is processed to 20cm×20cm×10cm.

[0117] The specific description of vibration signal collection is as follows:

[0118] The interaction between drill bit and rock is a complex mechanical process, and vibration is the physical information stimulated in this process, so the vibration signal is a carrier of comprehensive information. When collecting vibration signals, if we only consider the differences caused by different lithologies and ignore the influence of factors such as drilling parameters, it will inevitably limit the generalization ability of the lithology intelligent identification model.

[0119] In addition to the three drilling parameters of rotation speed, drilling speed and drill bit diameter, the drilling depth can also be taken into consideration because the homogeneity of each type of rock is different, the distribution of minerals is also different, and the collection conditions will be affected as the distance between the drill bit and the sensor changes.

[0120] Three levels were set for each factor of drill bit diameter, rotation speed, and drilling speed. Each type of rock was drilled in a 3×3×3 comprehensive experiment and the corresponding vibration signals were collected. The drilling depth will continue to deepen over time as the drill bit breaks the rock, so there is no need to set different level groups for this factor. However, in order to collect the vibration signals of the drill bit eating and stable drilling at the same time, based on the fact that the height of the drill bit used in the experiment is 10mm, the final depth of the drilling is set to 20mm, and the vibration signals at different hole depths are subsequently compared and analyzed. The design of the drilling experiment plan is shown in the following table:

[0121] Table 2 Level design of each factor for each lithology

[0122]

[0123] 27 experimental groups were designed for each type of rock type. Collecting vibration signals in multiple groups can not only effectively increase the samples required for model construction, but the diversity of the data can also improve the generalization of the model.

[0124] 3. Experimental Results

[0125] The characteristic analysis of the vibration signal is as follows:

[0126] 1. Signal waveform analysis

[0127] The vibration signal collected under the condition of sampling rate fs = 51200Hz is a high-frequency broadband signal, and each sampling point represents the triaxial vibration g at the interval time Δt = 1.953×10-5s. Therefore, the signal waveform is drawn with the acquisition time and signal amplitude as the horizontal and vertical coordinates, which can intuitively reflect the fluctuation process of the vibration signal when drilling into the rock.

[0128] In terms of the waveform characteristics of each type of lithology, the signal waveforms of shale, marl and limestone have small fluctuations, and the three have the largest rock mechanical parameters among the seven types of lithology, so the interaction between the hard rock and the drill bit produces oscillations; while the waveforms of feldspar sandstone and quartz sandstone are stable, and the amplitude is significantly lower than that of the other five types of lithology, because the rock mechanical parameters of the two are small and the rocks are relatively soft. However, granite, which has a medium rock mechanical parameter, has a relatively stable waveform and its signal amplitude has the maximum value among all types of rocks. In addition, although the rock mechanical parameters of granite and marble are similar in size, the signal amplitudes of the two show a large difference. Therefore, the waveform characteristics of the vibration signal are not only related to the rock mechanical parameters of the lithology itself, but also affected by factors such as rock structure. It is difficult to effectively distinguish different types of lithology based on the signal waveform alone, so it is also necessary to analyze the time domain characteristics and frequency domain characteristics of the signal.

[0129] 2. Time domain analysis

[0130] As a physical quantity of the rock drilling process, the vibration signal carries a wealth of information about the drilling status and rock characteristics. Information extraction requires the analysis and processing of the signal. To analyze the mathematical characteristics of the signal, the time domain, amplitude domain, frequency domain, and time-frequency domain are often used to describe it. The 2s observation scale is selected for analysis in each domain.

[0131] 2.1 Time domain drilling state characteristics

[0132] The characteristics of vibration signals can reflect two types of information: drilling status and rock type. First, the drilling status characteristics of vibration signals are analyzed.

[0133] Each type of rock is drilled under the conditions of rotation speed, drilling speed, and drill bit diameter at 3×3×3 levels, and the vibration signal of each borehole with a hole depth of 0 to 20 mm is collected. Taking the drilling state of granite as an example, the time domain characteristic information of the vibration signal under the influence of rotation speed, drilling speed, drill bit diameter and drilling depth is analyzed respectively.

[0134] Figure 1 The figure shows the time domain characteristics of the vibration signal of granite at a drilling depth of 20 mm at different speeds when the drill bit diameter is 30 mm and the drilling speed is 3 mm / min. The change in speed has a greater impact on the impact characteristics of the z-axis vibration signal, but the impact on the x-axis and y-axis vibration signals in the direction of drill bit rotation is reflected in the energy level, indicating that the change in speed affects the energy of drill bit cutting, resulting in a trend of impact characteristic changes in the feed direction.

[0135] Figure 2 The time domain characteristics of the vibration signal of granite at a drilling depth of 20mm under different drilling speeds when the drill bit diameter is 30mm and the speed is 300RPM. The change in drilling speed has no obvious effect on the impact characteristics of the vibration signal. The change in characteristics is mainly reflected in the energy level, and it has the overall characteristics that the greater the drilling speed, the greater the energy of the z-axis vibration signal, but the smaller the energy of the x-axis and y-axis vibration signals.

[0136] Figure 3 The figure shows the time domain characteristics of the vibration signal of granite at a drilling depth of 20 mm under different drill bit diameters at a rotation speed of 300 RPM and a drilling speed of 3 mm / min. The change in drill bit diameter has a certain influence on the energy level and impact characteristics of the vibration signal, and in general, the larger the drill bit diameter, the higher the energy level of the vibration signal and the more obvious the impact characteristics.

[0137] Figure 4The time domain characteristics of the vibration signals at 1mm, 10mm and 20mm are obtained by analyzing the drilling of granite under the conditions of 30mm drill bit, 300RPM speed and 3mm / min drilling speed. The influence of different drilling depths on the vibration signal is mainly reflected in the energy level. The energy level of the vibration signal increases continuously from the time when the drill bit just eats into the rock to the time when the drill bit is completely eaten into the rock. After the drill bit is completely eaten into the rock, the energy level begins to decrease as it continues to drill. It is worth noting that the decrease in energy level is related to the distance between the sensor and the drill bit to a certain extent.

[0138] Rotation speed, drilling speed, drill bit diameter and drilling depth have specific influences on the time domain characteristics of the vibration signal, so the vibration signal has characteristic information reflecting the drilling status.

[0139] 2.2 Rock type characteristics in the time domain

[0140] Each type of rock was broken under the drilling conditions of 30mm drill bit diameter, 400RPM rotation speed and 3mm drilling speed, and the vibration signal at the drilling depth of 20mm was analyzed, such as Figure 5 .

[0141] As can be seen from the figure, quartz sandstone and feldspar sandstone have low uniaxial compressive strength, so their energy level and impact characteristics are relatively low overall. Granite has high energy level and impact characteristics due to its structural characteristics of igneous rocks. Shale, marl and limestone have high uniaxial compressive strength, so they have high energy level and impact characteristics overall, and the energy level has a strong correlation with their uniaxial compressive strength. However, the impact characteristics have no obvious rules due to different rock structures. Compared with other types of rock types, marble has a medium energy level and impact characteristics.

[0142] 3. Frequency domain analysis

[0143] Frequency domain is one of the important ways to describe signal characteristics. By analyzing the frequency domain of the signal, we can obtain information such as the frequency structure and main frequency of the signal. Therefore, the frequency domain description of the vibration signal can further extract characteristic information about the drilling status and rock type from the signal.

[0144] 3.1 Frequency domain drilling state characteristics

[0145] Similar to the time domain feature analysis, the frequency domain feature analysis of the vibration signal also needs to be divided into two aspects: drilling state characteristics and rock type characteristics. First, the frequency domain drilling state characteristics are analyzed.

[0146] The vibration signals of granite under different drilling conditions were analyzed. Figure 6The power spectrum density of the vibration signal of granite at different speeds at 20mm drilling depth under the drilling conditions of 30mm drill bit diameter and 3mm / min drilling speed, and at different drilling speeds at 20mm drilling depth under the drilling conditions of 30mm drill bit diameter and 300RPM drilling speed. The frequency resolution of the power spectrum density is 40Hz, as shown below:

[0147]

[0148] In the formula, f s is the vibration signal sampling rate, its value is 51200 Hz; NFFT is the number of data points of Fourier transform, its value is 1280.

[0149] From the four power spectrum density graphs, it can be seen that as the rotation speed and drilling speed change, the amplitude of the power spectrum density of the vibration signal also changes, and the main frequency of the vibration signal between 6000-10000Hz is offset, indicating that the high frequency band of the vibration signal can reflect the drilling status information. Combining Tables 3 and 4, it can be seen that there is no significant difference in the frequency standard deviation at different rotation speeds and drilling speeds, indicating that the rotation speed and drilling speed have no significant effect on the dispersion of the power spectrum density amplitude, but the center of gravity frequency is greatly affected by the rotation speed and drilling speed, and the center of gravity frequency of the vibration signal decreases first and then increases with the increase of the rotation speed, and increases first and then decreases with the increase of the drilling speed, so the rotation speed and drilling speed also have different characteristics in the power spectrum density.

[0150] Table 3 Frequency domain characteristic indicators at different speeds

[0151]

[0152]

[0153] Table 4 Frequency domain characteristic indicators at different drilling speeds

[0154]

[0155] Figure 7 It is the power spectral density of vibration signals of granite at 20mm drilling depth under different drill bit diameters when the drilling speed is 300RPM and the drilling speed is 3mm / min, and at different drilling depths under the drilling conditions of 30mm drill bit diameter, 300RPM rotation speed and 3mm / min drilling speed.

[0156] It can be seen from the four power spectrum density diagrams that the drill bit diameter significantly affects the amplitude of the vibration signal power spectrum density between 6000-10000Hz. The change of drill bit diameter has no significant effect on the dispersion of the power spectrum density amplitude, but the center of gravity frequency gradually shifts to high frequency with the increase of drill bit diameter. Therefore, the vibration signal in the high frequency band also contains information characterizing the drilling status.

[0157] 3.2 Frequency domain rock type characteristics

[0158] Similar to the time domain feature analysis, the vibration signal of each type of rock at a drilling depth of 20 mm under the drilling conditions of a drill bit diameter of 30 mm, a rotation speed of 400 mm, a drilling speed of 3 mm / min, and a drilling depth of 20 mm was analyzed, and the power spectral density of the x-axis acceleration, y-axis acceleration, and z-axis acceleration when drilling each type of rock was obtained.

[0159] The power spectrum density of the vibration signal is closely related to the rock type, and the curve shape and main frequency distribution are affected by the rock structure and mineral composition. The frequency in the power spectrum density of the vibration signal that is significantly larger than the surrounding amplitude is selected as the main frequency. The difference between the centroid frequency and the frequency standard deviation of various lithologies is obvious, indicating that the power spectrum density of various lithologies has different curve shapes and main frequency amplitudes, which can be used as a basis for distinguishing lithologies.

[0160] Rock-breaking vibration is a physical phenomenon in which rocks move randomly in any direction in space under the action of the drill bit, so it has spatiality. The signals collected by the triaxial accelerometer represent vibrations in three directions, but their essence is the same physical phenomenon, so the main frequency distribution of the power spectrum density of the vibration signals in different directions of various lithologies is highly consistent. The power spectrum density of the vibration signals of each type of lithology is analyzed separately. The mineral grain size of granite and feldspar sandstone is the coarsest among the seven types of lithologies, so the power spectrum density shows that the main frequency amplitude in the medium and low frequency bands (3000-7000Hz) is prominent, while the main frequency in the high frequency band (7000-10000Hz) is not obvious or has no main frequency. The rock mechanical parameters of marl, limestone and shale are the largest, and the results of the time domain characteristic analysis show that the three have obvious impact characteristics, so the power spectrum density shows that the curve shape is steep and the main frequency peak is prominent. Marble has a high-frequency main frequency of 9650Hz with a prominent amplitude, which is considered to be related to its main mineral being fine-grained calcite. The quartz grain size of quartz sandstone is relatively coarse, so the main frequency amplitude in the medium and low frequency bands is relatively large.

[0161] It can be seen from Table 5 that the centroid frequency and frequency standard deviation of different lithologies are significantly different, indicating that the distribution of the power spectrum density curves of the vibration signals of various lithologies are significantly different, and the conditions for lithology identification are met.

[0162] Table 5 Frequency domain characteristic indicators of different rock types

[0163]

[0164]

[0165] Through the above embodiments and experimental examples, it can be seen that the present invention provides a method for extracting key indicative features of lithology from high-frequency vibration signals during drilling. The information in the time domain, frequency domain and time-frequency extracted by the method of the present invention is more closely related to lithology, and lithology classification can be better achieved. Therefore, the present invention has a good application prospect.

Claims

1. A method for extracting key indicative features of lithology from high-frequency vibration signals during drilling, characterized in that: The steps include: Step 1, real-time acquisition of high-frequency vibration signals generated during drilling; Step 2, preprocessing the high-frequency vibration signal during the drilling process; Step 3: extract the time domain, frequency domain and time-frequency information from the preprocessed vibration signal.

2. The feature extraction method according to claim 1, characterized in that: In step 1, information on the corresponding rotation speed, drilling speed and drill bit diameter when the high-frequency vibration signal during drilling is generated is also collected.

3. The feature extraction method according to claim 2, characterized in that: The rotation speed ranges from 200 to 400 RPM; the drilling speed ranges from 1 to 3 mm / min; the drill bit diameter ranges from 25 to 40 mm; and the drilling depth ranges from 0 to 20 mm.

4. The feature extraction method according to claim 1, characterized in that: In step 2, the specific method of preprocessing is selected from at least one of the following methods: pre-emphasis, wavelet threshold noise reduction and bandpass filtering.

5. The feature extraction method according to claim 4, characterized in that: In step 2, the wavelet threshold denoising selects the Sym8 wavelet basis function and performs 9-layer wavelet decomposition on the signal.

6. The feature extraction method according to claim 1, characterized in that: In step 3, before extracting the information in the time domain, frequency domain and time-frequency domain, it is verified that the vibration signal is a wide stationary random signal.

7. The feature extraction method according to claim 6, characterized in that: In step 3, the method for verifying that the vibration signal is a wide stationary random signal includes the following steps: Step a: Divide samples of different observation scales into N subsamples with specific frame lengths and frame shifts, and calculate the mean μ of each subsample i ; Step b, calculate the sample mean Record the successive μ in the subsample i <μ or μ i >μ is a run, and the total number of runs is R; i <μThe number of subsamples is m, satisfying μ i >The number of subsamples of μ is n; Step c, assume that the subsample mean μ i is a constant, and the normal distribution is used to test the acceptance domain. The test statistic Z is calculated according to the following formula: Where R is the numerator of the test statistic, μ R is the sample expectation, σ R is the sample standard deviation, and ~N(0,1) means normal distribution; where m is the number of subsamples satisfying u i < u, and n is the number of subsamples satisfying u i > u; Step d: set the significance level α = 0.05 for a two-sided test, and the acceptance domain of the null hypothesis is |Z| <Z α / 2 , when |Z|<1.96, accept the null hypothesis, subsample μ i is a constant, that is, the original random signal is a wide-stationary random signal.

8. The feature extraction method according to claim 1, characterized in that: In step 3, the time domain information is extracted by short-time Fourier transform.

9. A system for implementing the feature extraction method according to any one of claims 1 to 8, characterized in that: include: A signal acquisition module, used for real-time acquisition of high-frequency vibration signals generated during drilling; A preprocessing module, used for preprocessing the vibration high-frequency signal during the drilling process; The feature extraction module is used to extract the time domain, frequency domain and time-frequency information from the preprocessed vibration signal.

10. A computer-readable storage medium, characterized in that: Stored thereon is: a computer program for implementing the feature extraction method described in any one of claims 1-8.

Images