Abnormal Detection Method for Geological Drilling Process Based on Interval-Augmented Mahalanobis Distance

Abstract

The present invention discloses an abnormal detection method for the geological drilling process based on the spaced augmented Mahalanobis distance. First, obtain the sample sets of the drilling process under normal conditions and the samples to be detected, standardize them respectively to obtain the sample set X and the test set Y, and use X as the reference signal. Secondly, perform spaced augmentation on the samples in X and Y respectively, and calculate the mean vector and covariance matrix of X under different spacings and augmentation times. Thirdly, calculate the Mahalanobis distances between the training samples and the samples to be detected after each spaced augmentation. Finally, perform kernel density estimation on the Mahalanobis distance distribution of the training samples under different spacings and augmentation parameters, set the alarm limit, and compare the Mahalanobis distance of the samples to be detected with the alarm limit to determine whether an abnormality occurs. Finally, conduct experimental verification on the abnormal sticking of the drill, and prove that the present invention can effectively reduce the alarm delay and the false and missed alarm rates when an abnormality occurs, and improve the abnormal monitoring level of the complex geological drilling process.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control of geological exploration drilling processes, and particularly to an abnormal detection method for geological drilling processes based on interval-augmented Mahalanobis distance. Background Art

[0002] There are rich reserves of deep geological resources in China, with great development potential. To achieve the national resource and energy supply, geological resource exploration and development have become inevitable. Deep geological exploration is a very complex process with a large amount of randomness, fuzziness, and uncertainty, and it is impossible to directly observe the working conditions underground. Due to the prominent characteristics of high temperature, high pressure, and high disturbance underground, the drill bit often penetrates through mutation strata such as alternating hard and soft formations and rock-breaking zones, and drilling anomalies may occur at any time. This not only affects the progress of the entire project but also threatens the safety of life and property, causing huge drilling accidents and economic losses. Therefore, mastering the key information during the drilling process and accurately detecting the anomalies that occur during the drilling process in a timely manner have become important topics in geological drilling research.

[0003] Currently, the detection of drilling process anomalies can be briefly divided into two categories: (1) Model-based methods: By establishing different system mechanism models, training and identifying the signals during the drilling process. However, in the actual drilling process, the underground environment is complex and changeable, and the signals are weak, making it difficult to simulate the actual drilling process through a single model. (2) Data-driven methods: Most existing studies use the original drilling data or simple statistical features as inputs, and all the origin parameters or features are used as inputs to the classifier, resulting in a high complexity of the model and potentially poor performance for early anomaly detection.

[0004] Therefore, the weak abnormal signals in the complex geological drilling process and the difficulty in detection are technical problems that need to be urgently solved. Summary of the Invention

[0005] Aiming at the technical problems such as weak abnormal signals and difficult detection in the complex geological drilling process, the present invention provides an abnormal detection method for geological drilling processes based on interval-augmented Mahalanobis distance, which can effectively reduce the alarm delay and false and missed alarm rates when anomalies occur, and improve the abnormal monitoring level of the complex geological drilling process.

[0006] In the first aspect, an abnormal detection method for geological drilling processes based on interval-augmented Mahalanobis distance includes the following steps:

[0007] Obtain a section of drilling process data under normal conditions and the sample data to be detected

[0008] For perform data standardization to obtain the data set X as the reference signal, and use Statistical measure pair of characteristic parameters Perform data standardization on the corresponding parameters in it to obtain the sample set Y to be detected;

[0009] Perform interval augmentation on the data set X and the sample set Y to be detected to obtain the normal sample set X under different numbers of intervals λ and augmentation times L λ，L and the sample set Y to be detected λ，L , and calculate the covariance matrix ∑ λ，L of X λ，L and the mean vector μ λ，L ;

[0010] Calculate each sample x λ，L in the normal sample set X λ，L (g) and the Mahalanobis distance value of the mean vector μ λ，L under the covariance matrix ∑ λ，L ; Calculate each sample to be detected y λ，L in the sample set Y to be detected λ，L (k) and the Mahalanobis distance value d λ，L of the mean vector μ λ，L under the covariance matrix ∑ λ，L (k);

[0011] Use the Gaussian kernel density estimation method to estimate the distribution of the Mahalanobis distance values of all data in the normal sample set X under different numbers of intervals λ and augmentation times L λ，L ; Set the alarm limit A at a given confidence level λ，L ;

[0012] Compare the size of the Mahalanobis distance value d λ，L (k) and the alarm limit A λ，L . If the Mahalanobis distance value d λ，L (k) exceeds the alarm limit A λ，L , it is judged that an abnormality has occurred.

[0013] Furthermore, after the step of comparing the size of the Mahalanobis distance value d λ，L (k) and the alarm limit A λ，L , if the Mahalanobis distance value d λ，L (k) exceeds the alarm limit A λ，L , it is judged that an abnormality has occurred, and it further includes:

[0014] Calculate different evaluation indexes of the Mahalanobis distance value d λ，L (k) under different numbers of intervals λ and augmentation times L;

[0015] Calculate the local optimal value of the comprehensive index according to the TOPSIS method. The number of intervals λ and the augmentation times L corresponding to the local optimal value of the comprehensive index are the local optimal values of the number of intervals and the augmentation times, respectively.

[0016] Further, the abnormal drilling process includes: drill string bouncing, sticking, drill string breakage, well kick, and lost circulation.

[0017] Further, the specific process of interval augmentation is as follows:

[0018] According to different numbers of intervals λ and augmentation times L, the interval-augmented sample vector formed by the data points at the current moment includes the data samples at the current moment and the data samples at the first L sampling moments with the number of intervals being λ, and they are stacked to form an interval-augmented vector.

[0019] At time g, the interval-augmented vector x λ，L (g) in the normal sample set X is calculated as follows:

[0020] x λ，L (g) = [x(g) … x(g - λ(j - 1)) … x(g - λ(L - 1))]

[0021] where x(g) represents the g-th sample vector in the normal sample set X, x(g - λ(j - 1)) represents the (g - λ(j - 1))-th sample vector in the normal sample set X, x(g - λ(L - 1)) represents the (g - λ(L - 1))-th sample vector in the normal sample set X, and j represents the j-th augmentation.

[0022] At time k, the interval-augmented vector y λ，L (k) in the sample set Y to be detected is calculated as follows:

[0023] y λ，L (k) = [y(k) … y(k - λ(j - 1)) … y(k - λ(L - 1))]

[0024] where y(k) represents the k-th sample vector in the sample set Y to be detected, y(k - λ(j - 1)) represents the (k - λ(j - 1))-th sample vector in the sample set Y to be detected, y(k - λ(L - 1)) represents the (k - λ(L - 1))-th sample vector in the sample set Y to be detected, and j represents the j-th augmentation.

[0025] Further, the calculation formula of the Mahalanobis distance is as follows:

[0026] At time g, the Mahalanobis distance value λ，L in the normal sample set X is calculated as:

[0027]

[0028] At time k, the sample set Y to be detected λ，L The Mahalanobis distance value d λ，L (k) is calculated as follows:

[0029]

[0030] where ∑ λ，L -1 represents the inverse matrix of the covariance matrix ∑ λ，L Find the inverse matrix.

[0031] Furthermore, the use of the Gaussian kernel density estimation method to estimate the distribution of the Mahalanobis distance values of all data in the normal sample set X under different numbers of intervals λ and augmentation times L λ，L The calculation formula is as follows: The calculation formula is as follows:

[0032]

[0033] where x is the independent variable, representing the continuous augmented Mahalanobis distance value, h is the bandwidth selected for kernel density estimation, N1 represents the number of samples in the normal sample set X, [N1 + λ(1 - L)] is the number of samples in the normal sample set X after interval augmentation, i represents the ordinal number of the normal data sample, is the Mahalanobis distance of any normal sample data after interval augmentation, k() is the Gaussian kernel function used, and the calculation formula is as follows:

[0034]

[0035] where exp represents the exponential function.

[0036] Furthermore, the setting of the alarm limit A at a given confidence level λ，L , includes:

[0037] Given the confidence level δ, solve the alarm limit A through the probability density integral formula λ，L , and the probability density integral formula is as follows:

[0038]

[0039] In a second aspect, the present invention provides a geological drilling process anomaly detection device based on interval-augmented Mahalanobis distance, including the following units:

[0040] A data acquisition module for acquiring drilling process data in a normal state and sample data to be detected

[0041] A standardization module for Performing data standardization to obtain the data set X as a reference signal, and using Statistical metrics pair of characteristic parameters Perform data standardization on the corresponding parameters in it to obtain the sample set Y to be detected;

[0042] Interval augmentation module, used to perform interval augmentation on the data set X and the sample set Y to be detected, and obtain the normal sample set X under different intervals λ and augmentation times L λ，L and the sample set Y to be detected λ，L , and calculate the covariance matrix ∑ of X λ，L under the corresponding interval augmentation times λ，L and the mean vector μ λ，L ;

[0043] Mahalanobis distance calculation module, used to calculate the Mahalanobis distance value of each sample x λ，L in the normal sample set X λ，L (g) and the mean vector μ λ，L under the covariance matrix ∑ λ，L ; It is also used to calculate the Mahalanobis distance value d λ，L of each sample y λ，L (k) to be detected in the sample set Y to be detected and the mean vector μ λ，L under the covariance matrix ∑ λ，L ; λ，L (k);

[0044] Alarm limit setting module, used to estimate the distribution of Mahalanobis distance values of all samples in the normal sample set X under different intervals λ and augmentation times L using the Gaussian kernel density estimation method λ，L and set the alarm limit A at a given confidence level ; λ，L ;

[0045] Abnormality judgment module, used to compare the size of the Mahalanobis distance value d λ，L (k) and the alarm limit A λ，L . If the Mahalanobis distance value d λ，L (k) exceeds the alarm limit A λ，L , it is judged that an abnormality has occurred.

[0046] In a third aspect, an electronic device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the steps of the above-mentioned abnormal detection method for the geological drilling process based on interval-augmented Mahalanobis distance.

[0047] In a fourth aspect, a storage medium stores a computer program, and when the computer program is executed by a processor, it implements the steps of the above-mentioned abnormal detection method for the geological drilling process based on interval-augmented Mahalanobis distance.

[0048] ​The technical solution provided by the present invention has the following beneficial effects:

[0049] The present invention provides a method for detecting anomalies in the geological drilling process based on the interval-augmented Mahalanobis distance. First, a sample set of the drilling process under normal conditions is obtained and a sample set to be detected Data standardization is performed on them to obtain the sample set X and the test set Y respectively, and X is used as the reference signal. Secondly, interval augmentation is performed on the samples in X and Y respectively, and the mean vector and covariance matrix of X under different intervals and augmentation parameters are calculated. Thirdly, the Mahalanobis distances of the training samples and the samples to be detected are calculated under the given intervals and augmentation parameters. Finally, kernel density estimation is performed on the distribution of the Mahalanobis distances of the training samples under different interval augmentation parameters, and an alarm limit is set. The calculated Mahalanobis distance of the sample to be detected is compared with the alarm limit to distinguish whether an anomaly has occurred. The present invention can improve the detection ability and detection accuracy of weak anomaly signals in the drilling process, reduce the detection alarm delay, and ensure the safe and stable operation of the drilling process. Description of the Drawings

[0050] The following takes the abnormal drill string bounce in the drilling process as an example and further illustrates the present invention in combination with the drawings. In the drawings:

[0051] Figure 1 is a flowchart of a method for detecting anomalies in the geological drilling process based on the interval-augmented Mahalanobis distance according to the present invention;

[0052] Figure 2 is a time series diagram of the drilling parameters used for detecting abnormal drill string bounce according to the present invention. Among them, the time series of normal data is in the range of t = 0s to 1500s, and the time series of the characteristic parameters under abnormal drill string bounce is in the range of t = 1501s to 3500s;

[0053] Figure 3 is a comparison diagram of the fitting effects of the kernel density estimation of the Mahalanobis distance of normal samples, the chi-square distribution, and the normal distribution under the optimal interval (λ = 3) and augmentation times (L = 7) when applied to the detection of abnormal drill string bounce according to the present invention;

[0054] Figure 4 is a surface diagram of the comprehensive index under different parameters λ and L. The ellipse is the selected optimal comprehensive index (λ = 3, L = 7);

[0055] Figure 5 From top to bottom are the T of PCA 2, The time series diagrams for alarm monitoring of two indicators, SPE, the time series diagram for alarm monitoring of the standard Mahalanobis Distance (MD), and the time series diagram of the Mahalanobis Distance under the optimal Interval Augmented (IA) parameters. Among them, the broken line in the figure is the time series under different monitoring indicators, and the virtual straight line is the alarm threshold. The data points where the monitored indicator values exceed the alarm threshold are identified as abnormal data points;

[0056] Figure 6 is the structural diagram of an abnormal detection device for geological drilling process based on interval augmented Mahalanobis distance of the present invention. Specific embodiments

[0057] For a clearer understanding of the technical features, objectives, and effects of the present invention, a method for a specific drilling anomaly - bit bounce anomaly - is now described, and the specific embodiments of the present invention are described in detail with reference to the accompanying drawings.

[0058] Figure 1 Overall, it shows the implementation steps of the present invention, providing a method for abnormal detection of geological drilling process based on interval augmented Mahalanobis distance. The specific steps are as follows:

[0059] S1: Obtain a section of drilling process data under normal conditions (training set) and the data of the sample to be detected (test set); Perform data standardization on to obtain the data set X as the reference signal, and use the statistical metrics of the characteristic parameters in to perform data standardization on the corresponding parameters in the data to be detected

[0060] The bit bounce anomaly is mainly caused by encountering gravel layers, hard and soft interlayers, and uneven limestone formations during drilling, resulting in uneven forces on the cone bits. The bit bounce anomaly can mainly be reflected by changes in torque, hook load, and rotational speed. Specifically, the fluctuation ranges of torque and hook load gradually increase, and for the rotational speed, since it may break away from the formation, the signal will increase by a certain amplitude. Therefore, in the embodiments of the present invention, the changes in three characteristic quantities, torque, hook load, and rotational speed, are selected to reflect whether the bit bounce anomaly occurs. Specifically, the above drilling engineering parameters can be measured by sensors.

[0061] It should be noted that the method of the present invention is applicable to various abnormal detections during the drilling process, such as sticking, drill pipe breakage, well kick, and lost circulation, and is not limited to the bit bounce anomaly.

[0062] Specifically, for the drilling process data under normal conditions For Perform Z-Score normalization on the dataset X to obtain the reference signal. Using the mean and standard deviation of the characteristic parameters in perform Z-Score normalization on the corresponding parameter data in as the sample Y to be detected. The Z-Score normalization methods for the training set and the data to be detected

[0063]

[0064]

[0065] are as follows: The selected characteristic parameters are in three dimensions. Any feature vector N1 represents the number of samples in the training set (500 data are selected in the present invention). The mean matrix The mean vector of any feature vector represents the mean of the j-th characteristic parameter data in the training set The standard deviation matrix σ = [σ1 σ2 σ3]. The standard deviation vector of any characteristic parameter represents the standard deviation of the j-th characteristic parameter data in the training set . Any feature vector N2 represents the number of samples to be detected (3500 data are selected in the present invention).

[0066] In addition, in other embodiments, the data normalization method can also select min-max normalization, log function conversion, and fuzzy quantization method.

[0067] Figure 2 shows the original data of the drilling process selected in the present invention, where the normal data time series is in the range from t = 0s to 1500s, and the time series of the three characteristic parameters under the abnormal drill string bounce is in the range from t = 1501s to 3500s.

[0068] S2: Perform interval augmentation on the dataset X and the sample set Y to be detected to obtain the normal sample set X λ，L and the sample set Y λ，L to be detected under different numbers of intervals λ and augmentation times L, and calculate the covariance matrix ∑ λ，L and the mean vector μ λ，L of X λ，L under the corresponding interval augmentation times;

[0069] The interval augmentation of the dataset X is as follows:

[0070]

[0071] Among them, λ and L represent the number of interval samples and the number of augmentations respectively, and N1 is the number of samples in the normal sample set X.

[0072] At the test set time instant k, the interval-augmented sample y λ，L (k) is calculated as follows:

[0073] y λ，L (k) = [y(k)…y(k - λ(j - 1))…y(k - λ(L - 1))]

[0074] Among them, y(k) represents the k-th sample vector in the sample set Y to be detected, y(k - λ(j - 1)) represents the (k - λ(j - 1))-th sample vector in the sample set Y to be detected, y(k - λ(L - 1)) represents the (k - λ(L - 1))-th sample vector in the sample set Y to be detected, and j represents the j-th augmentation.

[0075] S3: Calculate the Mahalanobis distance value of each sample x λ，L in the normal sample set X λ，L (g) and the mean vector μ λ，L (training samples) under the covariance matrix ∑ λ，L ; Calculate the Mahalanobis distance value d λ，L of each sample y λ，L (k) to be detected in the sample set Y to be detected and the mean vector μ λ，L (test samples) under the covariance matrix ∑ λ，L ; λ，L (k);

[0076] The calculation method of the Mahalanobis distance is as follows:

[0077] At time g, the Mahalanobis distance value λ，L in the normal sample set X is calculated as:

[0078]

[0079] Among them, ∑ λ，L -1 represents the inverse matrix of the covariance matrix ∑ λ，L ; μ i is the mean of the data in the i-th column of the interval-augmented sample set X λ，L .

[0080] At time k, the Mahalanobis distance value d λ，L in the sample set Y to be detected λ，L (k) is calculated as:

[0081]

[0082] Define the covariance between any two dimensions C i and C j as follows:

[0083] Cov(C i , C j ) = E[(C i - E(C i ))(C j - E(C j ))]

[0084] where E(C i ) and E(C j ) are the expected values of dimensions C i and C j respectively. Then the covariance matrix ∑ λ，L can be calculated as:

[0085]

[0086] S4: Use the Gaussian kernel density estimation method to estimate the distribution of Mahalanobis distance values of all data in the normal sample set X λ，L under different intervals λ and augmentation times L, and set the alarm limit A λ，L at a given confidence level;

[0087] First, use the Gaussian kernel density estimation method to estimate the distribution of Mahalanobis distance of the normal data sample set X λ，L under different intervals λ and augmentation times L in step S2, and then set the alarm limit A λ，L at a given confidence interval. The Gaussian kernel density function is estimated as follows:

[0088]

[0089] where x is the independent variable, representing the continuous augmented Mahalanobis distance value, h is the bandwidth selected for kernel density estimation, here it is the default value 5.0474, N1 represents the number of samples in the normal sample set X, [N1 + λ(1 - L)] is the number of samples in the normal sample set X after interval augmentation (in this embodiment, 500 normal sample vectors are selected), i represents the ordinal number of the normal data sample, is the Mahalanobis distance of the i-th data in the normal sample set X λ，L after interval augmentation, and k() is the Gaussian kernel function adopted, and the calculation method is as follows:

[0090]

[0091] Among them, exp represents the exponential function. After estimating the Mahalanobis distance density distribution of normal samples at different intervals and augmentation times, given the confidence level δ (δ = 99% is set in the present invention), the alarm limit A is solved through the probability density integration formula λ，L , and the probability density integration formula is as follows:

[0092]

[0093] Figure 3 Fig. shows the comparison diagram of the fitting effects of the kernel density estimation of the Mahalanobis distance of normal samples, the chi-square distribution, and the normal distribution estimation under the optimal interval (λ = 3) and augmentation times (L = 7) in the embodiments of the present invention. The dotted line is the kernel density estimation density curve, the straight curve is the chi-square distribution density curve, and the dotted line is the normal distribution density curve

[0094] S5: Compare d λ，L (k) with the alarm limit A λ，L to judge whether the detection is abnormal by comparing their magnitudes

[0095] Under different numbers of intervals and augmentation times, it is stipulated that when the value of d λ，L (k) exceeds the alarm limit A λ，L , it is considered that a abnormal drill jump occurs, and an alarm is given at the (k + 1)th moment

[0096] After step S5, it further includes:

[0097] Verification indicators and comprehensive

[0098] The four verification indicators of the test set are as follows:

[0099]

[0100]

[0101]

[0102] Among them, Acc is the correct rate, MAR is the missed alarm rate, FAR is the false alarm rate, TP is the true positive example, TN is the true negative example, FP is the false positive example, and FN is the false negative example. In addition, the data sample to be detected Y ∈ R 3500×3 , the first 1500 samples are normal data, and the last 2000 samples are abnormal data. It is stipulated that the time period from the 1501s time point to a certain time point when an alarm continuously occurs for 10s is the detection delay (DD)

[0103] In order to comprehensively balance the indicators and select the optimal interval augmentation, the technique for order preference by similarity to an ideal solution (TOPSIS) is used here to unify the above four verification indicators into 1 indicator. By selecting different numbers of intervals λ and augmentation times L, calculate d under different λ and Lλ，L (k) for different indicators and parameter tuning. The optimization intervals selected in the embodiments of the present invention are the number of intervals λ(a)=1, 2, 3, 4, 5, and L(b)=1 to 30. The steps of the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) are as follows:

[0104] 1. Convert all indicators into benefit indicators;

[0105] Let the indicator matrix (There are 150 combinations of λ and L) Since the smaller the values of the indicators MAR, FAR, and DD, the better the effect, these three indicators are converted into benefit indicators as follows:

[0106]

[0107] where i ∈ 1 to 150, representing the number of rows of the indicator matrix, and j ∈ 1 to 4, representing the number of columns of the indicator matrix.

[0108] 2. Standardize the positive matrix to obtain the standardized matrix Ind:

[0109] Ind = [Acc MAR FAR DD]

[0110] where

[0111] 3. Calculate the comprehensive evaluation index (Comprehensive Indicators, CI) using Ind

[0112] For the comprehensive evaluation index CI(i) under any interval augmentation parameter, the calculation is as follows:

[0113]

[0114] where That is, find the comprehensive index CI under a certain interval augmentation parameter (λ and L) such that the distance from CI to the optimal indicator in each dimension is the closest. The interval augmentation parameter corresponding to the local optimal value of the comprehensive index CI is the local optimal value of the interval augmentation parameter.

[0115] Figure 4 Shows the surface plot of the comprehensive index under different parameters λ and L, where the ellipse is the selected optimal comprehensive index (obtained at λ = 3 and L = 7).

[0116] To demonstrate the effectiveness of the proposed method, the embodiments of the present invention are compared with other methods (including T of PCA (Principal Component Analysis)) 2Compare with the SPE (Squared Prediction Error) index and the standard Mahalanobis distance Figure 5 From top to bottom are the T of PCA 2 The alarm monitoring time series diagrams of the two indicators of T and SPE, the alarm monitoring time series diagram of the standard Mahalanobis distance (MD), and the time series diagram of the Mahalanobis distance under the parameters of the locally optimal interval augmentation (Interval Augmented, IA). Among them, the broken line in the figure is the time series under different monitoring indicators, and the virtual straight line is the alarm limit. The data points whose monitoring indicator values exceed the alarm limit are identified as abnormal data points. And the comparison of each evaluation index of different methods is given in Table 1. Compared with other methods, the proposed method has good overall results.

[0117] Table 1 Comparison of indexes of different methods

[0118]

[0119] Figure 5 Overall, it shows an abnormal detection device for geological drilling process based on interval-augmented Mahalanobis distance, which is used to implement the above-mentioned abnormal detection method for geological drilling process. The device specifically includes the following modules:

[0120] Data acquisition module 1, which is used to acquire the drilling process data under normal conditions And the sample data to be detected

[0121] Standardization module 2, which is used to Perform data standardization on to obtain the data set X as the reference signal, and use the mean and standard deviation of the characteristic parameters in to perform data standardization on the corresponding parameters in as the sample set Y to be detected; In the In the corresponding parameters in

[0122] Interval augmentation module 3, which is used to perform interval augmentation on the data set X and the sample set Y to be detected to obtain the normal sample set X and the sample set Y to be detected under different intervals λ and augmentation times L λ，L And the sample set Y to be detected λ，L , and calculate the covariance matrix ∑ λ，L Of X λ，L And the mean vector μ λ，L ;

[0123] Mahalanobis distance calculation module 4, which is used to calculate the Mahalanobis distance value of each sample x λ，L In the normal sample set X λ，L (g) and the mean vector μ λ，L Under the covariance matrix ∑ λ，L ​It is also used to calculate the sample set Y to be detected λ，L each sample y to be detected λ，L (k) and the mean vector μ λ，L under the covariance matrix ∑ λ，L the Mahalanobis distance value d λ，L (k);

[0124] An alarm limit setting module 5, which is used to estimate the distribution of the Mahalanobis distance values of all samples in the normal sample set X under different intervals λ and augmentation times L by using the Gaussian kernel density estimation method λ，L and set the alarm limit A at a given confidence level ; λ，L ;

[0125] An anomaly judgment module 6, which is used to compare the size of the Mahalanobis distance value d λ，L (k) and the alarm limit A λ，L , if the Mahalanobis distance value d λ，L (k) exceeds the alarm limit A λ，L , it is determined that an anomaly has occurred

[0126] In addition, this embodiment also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the in-hole jumping drilling anomaly detection method based on interval-augmented Mahalanobis distance are implemented

[0127] In addition, this embodiment also provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the anomaly detection method for the geological drilling process based on interval-augmented Mahalanobis distance are implemented

[0128] It should be noted that in this article, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or system. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or system including that element

[0129] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. Among the several apparatus unit claims listed, several of these apparatuses may be embodied by the same hardware item. The use of the words first, second, and third does not denote any order and these words may be construed as identifiers

[0130] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. An abnormal detection method for geological drilling process based on interval-augmented Mahalanobis distance, characterized in that, It includes the following steps: Obtain the drilling process data under normal conditions and the sample data to be detected Pair Data normalization is performed on to obtain the data set X, which serves as the reference signal. Then, using the statistical metrics of the feature parameters in data normalization is performed on the corresponding parameters in to obtain the sample set Y to be detected; Perform interval augmentation on the dataset X and the sample set Y to be detected to obtain the normal sample set X under different numbers of intervals λ and augmentation times L λ，L and the sample set Y to be detected λ，L , and calculate the covariance matrix ∑ λ，L of X λ，L and the mean vector μ λ，L ; Calculate the normal sample set X λ，L for each sample x λ，L (g) and the mean vector μ λ，L under the covariance matrix ∑ λ，L of the Mahalanobis distance value Calculate the sample set Y to be detected λ，L for each sample f to be detected λ，L (k) and the mean vector μ λ，L under the covariance matrix ∑ λ，L of the Mahalanobis distance value d λ，L (k); Estimate the distribution of Mahalanobis distance values of all data in the normal sample set X under different numbers of intervals λ and augmentation times L using the Gaussian kernel density estimation method λ，L in Set the alarm limit A at a given confidence level λ，L ; Compare the Mahalanobis distance value d λ，L (k) with the alarm limit A λ，L in terms of magnitude. If the Mahalanobis distance value d λ，L (k) exceeds the alarm limit A λ，L , an abnormality is determined to have occurred.

2. The abnormal detection method for the geological drilling process based on the spaced augmented Mahalanobis distance according to claim 1, characterized in that After comparing the Mahalanobis distance value d λ，L (k) with the alarm limit A λ，L to determine their magnitudes, if the Mahalanobis distance value d λ，L ]](k) exceeds the alarm limit A(k) exceeds the alarm limit A λ，L , after the step of determining that an abnormality has occurred, it further includes: Calculate the Mahalanobis distance value d under different numbers of intervals λ and augmentation times L λ，L (k) of different evaluation indicators; Calculate the local optimal value of the comprehensive index according to the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS). The number of intervals λ and the augmentation times L corresponding to the local optimal value of the comprehensive index are the local optimal values of the number of intervals and the augmentation times.

3. The method according to claim 1, wherein the abnormal drilling process includes: Hammering, sticking, drill pipe breakage, well kick and lost circulation.

4. The abnormal detection method for the geological drilling process based on the spaced augmented Mahalanobis distance according to claim 1, wherein The specific process of interval augmentation is as follows: According to the different values of the number of intervals λ and the augmentation times L, the interval-augmented sample vector formed by the data points at the current moment includes the data sample at the current moment and the data samples at the first L sampling moments with the number of intervals being λ, which are stacked to form an interval-augmented vector. At time g, the spaced augmented vector x in the normal sample set X λ，L (g) is calculated as follows: x λ，L (g) = [x(g) … x(g - λ(j - 1)) … x(g - λ(L - 1))] Among them, x(g) represents the g-th sample vector in the normal sample set X, x(g - λ(j - 1)) represents the (g - λ(j - 1))-th sample vector in the normal sample set X, x(g - λ(L - 1)) represents the (g - λ(L - 1))-th sample vector in the normal sample set X, and j represents the j-th augmentation. At time k, the spaced augmented vector y λ，L (k) is calculated as follows: y λ，L y(k) = [y(k) … y(k - λ(j - 1)) … y(k - λ(L - 1))] Among them, y(k) represents the k-th sample vector in the sample set Y to be detected, y(k - λ(j - 1)) represents the (k - λ(j - 1))-th sample vector in the sample set Y to be detected, y(k - λ(L - 1)) represents the (k - λ(L - 1))-th sample vector in the sample set Y to be detected, and j represents the j-th augmentation.

5. The abnormal detection method for the geological drilling process based on the spaced augmented Mahalanobis distance according to claim 1, characterized in that The Mahalanobis distance calculation formula is as follows: At time g, the normal sample set X λ，L The Mahalanobis distance value is calculated as follows: At time k, the sample set Y to be detected λ，L The Mahalanobis distance value d λ，L (k) is calculated as follows: Among them, ∑ λ，L -1 represents the inverse matrix of the covariance matrix ∑ λ，L obtained by taking the inverse matrix.

6. The abnormal detection method for the geological drilling process based on the spaced augmented Mahalanobis distance according to claim 1, characterized in that Estimating the distribution of Mahalanobis distance values of all data in the normal sample set X under different numbers of intervals λ and augmentation times L using the Gaussian kernel density estimation method λ，L is as follows The calculation formula is as follows: Among them, x is the independent variable, representing the continuous augmented Mahalanobis distance value, h is the bandwidth selected for kernel density estimation, N1 represents the number of samples in the normal sample set X, [N1 + λ(1 - L)] is the number of samples after the normal sample set X is augmented at intervals, and i represents the ordinal number of the normal data samples. is the Mahalanobis distance of any normal sample data after interval augmentation, and k() is the Gaussian kernel function used, and the calculation formula is as follows: Among them, exp represents the exponential function.

7. The abnormal detection method for the geological drilling process based on the interval-augmented Mahalanobis distance according to claim 6, wherein Setting the alarm limit A at a given confidence level λ，L , including: Given the confidence level δ, solve for the alarm limit A using the probability density integration formula λ，L , and the probability density integration formula is as follows:

8. An abnormal detection device for geological drilling process based on interval-augmented Mahalanobis distance, characterized in that, It includes the following units: A data acquisition module for acquiring drilling process data under normal conditions and sample data to be detected A standardization module for to perform data standardization to obtain a data set X as a reference signal, and use the statistical metrics of the feature parameters in to perform data standardization on the corresponding parameters in it as a sample set Y to be detected; Interval augmentation module, which is used to perform interval augmentation on the dataset X and the sample set Y to be detected, and obtain the normal sample set X under different intervals λ and augmentation times L λ，L and the sample set Y to be detected λ，L , and calculate the covariance matrix ∑ λ，L of X λ，L and the mean vector μ λ，L ; Mahalanobis distance calculation module, used to calculate the Mahalanobis distance value of each sample x λ，L in the normal sample set X λ，L (g) from the mean vector μ λ，L under the covariance matrix ∑ λ，L ; It is also used to calculate the Mahalanobis distance value d λ，L (k) of each sample y λ，L in the sample set Y to be detected λ，L (k) from the mean vector μ λ，L under the covariance matrix ∑ λ，L (k). An alarm limit setting module for estimating the distribution of the Mahalanobis distance values of all samples in the normal sample set X under different intervals λ and augmentation times L by using the Gaussian kernel density estimation method λ，L in the normal sample set X Set the alarm limit A at a given confidence level λ，L ; Anomaly judgment module, used to compare the Mahalanobis distance value d λ，L (k) with the alarm limit A λ，L in terms of magnitude. If the Mahalanobis distance value d λ，L (k) exceeds the alarm limit A λ，L , it is judged that an anomaly has occurred.

9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the abnormal detection method for the geological drilling process based on the interval-augmented Mahalanobis distance described in any one of claims 1 - 7.

10. A storage medium, on which a computer program is stored, characterized in that, When the computer program is executed by the processor, it implements the steps of the abnormal detection method for the geological drilling process based on the interval-augmented Mahalanobis distance described in any one of claims 1 - 7.

Images