A method for locating wellbore defects

Abstract

The present invention discloses a method for locating a wellbore defect, including synchronous timing of a defect detection device, an inertial navigation system, and a three-axis magnetometer; obtaining defect detection data, inertial navigation data, and three-axis magnetometer data, and aligning them by time; establishing wellbore ultrasonic / magnetic leakage imaging, and extracting feature information and corresponding positioning data in the wellbore imaging; obtaining a data set and a position set of n measurements; selecting any position set as a basis, and obtaining a mapping relationship between other position sets and the position set through feature recognition and key point matching methods, and interpolating to obtain a complete mapping function; correcting the positioning data of the defect feature according to the inverse function of the mapping function to obtain the final positioning data. The present invention can directly use the magnetic field information in the wellbore for defect positioning, solve the problem of error accumulation of the gyroscope, and correct the positioning data through feature extraction and key point matching, thereby improving the positioning accuracy of the wellbore defect.

Description

Technical Field

[0001] The invention relates to a wellbore defect positioning method and belongs to the technical field of defect detection. Background Art

[0002] Current defect detection technologies include ultrasonic or magnetic flux leakage detection, which uses the change of magnetic induction lines at the wellbore defect to detect and quantify the nature and quality of the defect. Existing defect positioning includes circumferential positioning and depth positioning. However, for the position of the defect that can be located at a specific point on the section, the positioning result has a large error due to the correction of the accelerometer. Summary of the invention

[0003] In order to overcome the defects in the prior art, the present invention aims to provide a method for locating wellbore defects.

[0004] The present invention provides a technical solution to solve the above technical problems: a method for locating a wellbore defect, comprising the following steps:

[0005] Step A, synchronously timing the defect detection device, the inertial navigation system, and the three-axis magnetometer; placing the defect detection device into the wellbore, performing defect detection on the wellbore, acquiring defect detection data, inertial navigation data, and three-axis magnetometer data, and aligning them according to time;

[0006] Step B, correcting the gyroscope data based on the local geomagnetic field, the trajectory data of the wellbore, the three-axis magnetometer data, and the shielding coefficient data;

[0007] Step C, combining the gyroscope and accelerometer information in the inertial navigation system with the data of the scanning device to establish the well perimeter ultrasonic / magnetic leakage imaging, and extracting the characteristic points in the well perimeter imaging to obtain a characteristic point set;

[0008] Step D: Repeat steps AC several times to obtain a data set I of n measurements. n =(x n ,y n ,v n ) and the location set Feature point set;

[0009] Step E: Select any position set A a As the basis, the mapping function between other position sets and the basis is obtained through segmentation point selector and key point matching based on improved LightGlue algorithm;

[0010] Step F: Correct the wellbore imaging map according to the inverse function of the mapping function to obtain a corrected wellbore imaging map.

[0011] A further technical solution is that the defect detection device is an ultrasonic or magnetic leakage detection device.

[0012] A further technical solution is that the feature extraction in step C adopts SIFT algorithm.

[0013] A further technical solution is that the number of directional gradients of the SIFT algorithm is 8 and the sub-region grid is 4×4.

[0014] A further technical solution is that the specific process of step E includes:

[0015] a. Feature points are segmented through segmentation point selector;

[0016] b. Input the segmented feature points into the LightGlue network improved by hybrid position encoding, output the matching points, and then use the algorithm to fit them to get the fitting curve;

[0017] c. Draw the total distortion curve according to the fitting curve, and then calculate the inverse function and sub-pixel interpolation to obtain the mapping function.

[0018] A further technical solution is that the attention score calculation formula of the LightGlue network improved by hybrid position encoding is:

[0019]

[0020] Where: Q, K, V are the query matrix, key matrix and value matrix absolutely encoded by x coordinates; RoPE(Q y )、RoPE(K y ) is the query matrix and key matrix encoded by y-coordinate rotation.

[0021] A further technical solution is that the relationship of the mapping function in step E is:

[0022]

[0023] Where: A is the location set; f a,n A a To A n The mapping function of .

[0024] A further technical solution is that the correction formula in step F is:

[0025]

[0026] Where: A is the location set; f a,n A a To A n The mapping function of .

[0027] The present invention has the following beneficial effects: the present invention can directly use the magnetic field information in the wellbore to locate defects, avoid the interference of the geological environment, solve the problem of error accumulation of the gyroscope, and at the same time correct the positioning data through feature extraction and key point matching, thereby improving the positioning accuracy of wellbore defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a schematic diagram of the feature extraction and key point matching process;

[0029] Figure 2 It is the feature extraction and key point matching graph;

[0030] Figure 3 This is the structural diagram of the LightGlue network. DETAILED DESCRIPTION

[0031] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] A method for locating a wellbore defect of the present invention comprises the following steps:

[0033] Step A, synchronously timing the defect detection device, the inertial navigation system, and the three-axis magnetometer; placing the defect detection device into the wellbore, performing defect detection on the wellbore, acquiring defect detection data, inertial navigation data, and three-axis magnetometer data, and aligning them according to time;

[0034] Step B, correcting the gyroscope data based on the local geomagnetic field, the trajectory data of the wellbore, the three-axis magnetometer data, and the shielding coefficient data;

[0035] Step C, combining the gyroscope and accelerometer information in the inertial navigation system with the data of the scanning device to establish the well perimeter ultrasonic / magnetic leakage imaging, and using the SIFT algorithm (the number of directional gradients of the SIFT algorithm is 8, the sub-region grid is 4×4, and for each feature point, the feature descriptor length is 4×4×8=128) to extract feature points in the well perimeter imaging to obtain a feature point set;

[0036] Step D: Repeat steps AC several times to obtain a data set I of n measurements. n =(x n ,y n ,v n ) and the location set Feature point set;

[0037] Step E: Select any position set Aa As the basis, the mapping function between other position sets and the basis is obtained through segmentation point selector and key point matching based on improved LightGlue algorithm;

[0038] Step E1, the feature points are segmented by a segmentation point selector;

[0039] Segmentation point selector, the feature point set extracted by SIFT algorithm is defined as a constant BatchSize; select the first BatchSize points in the feature point set sorted by x coordinate, calculate their horizontal coordinates, calculate the average value, and take all points with horizontal coordinates less than this value into the LightGlue network;

[0040] Step E2, input the segmented feature points into the LightGlue network improved by hybrid position encoding, output the matching points, and then use the algorithm to fit them to obtain the fitting curve;

[0041] The original LightGlue network uses rotation encoding. This encoding method is a relative position method, which is particularly suitable for the image matching field to which the original algorithm is applicable. However, in the present invention, the image may have different characteristics at different absolute positions, for example: the circumferential position distortion is smaller than the axial position distortion, and the closer to the right side of the image (the closer to the bottom of the well), the greater the relative position distortion. At the same time, the image in the present invention has large differences in features in the x and y directions; the scale in the x direction (depth direction) is large, which is affected by the absolute position, and is significantly distorted by the error of the inertial navigation system; while the scale in the y direction (circumferential direction of the wellbore) is small, and has uniform characteristics on the circumference, which is not affected by the absolute position. Therefore, the present invention proposes a new hybrid encoding method, which can focus on introducing the absolute position in the x direction and the relative position information in the y direction into the transformer network.

[0042] The hybrid position coding combines the absolute position coding and the relative position coding, which superimposes or uses the information of the two together;

[0043] The attention score calculation formula of the LightGlue network improved by hybrid position encoding is:

[0044]

[0045] Where: Q, K, V are the query matrix, key matrix and value matrix absolutely encoded by x coordinates; RoPE(Q y )、RoPE(K y ) is the query matrix and key matrix encoded by y-coordinate rotation; d is the length of the query vector Qy;

[0046] in,

[0047] Q=(α+PE absolute )M Q

[0048] K=(α+PE absolute )M K

[0049] V=(α+PE absolute )M V

[0050]

[0051] Q y =αM Q

[0052] K y =αM K

[0053] Step E3, drawing a total distortion curve according to the fitting curve, and then calculating the inverse function and sub-pixel interpolation to obtain a mapping function;

[0054] Since the position set of the nth measurement Then there is a mapping function f n So that:

[0055] A n =f n (B)

[0056] If B is estimated by the mean of a known measurement data set, then the mean is the mean of the corresponding points after mapping:

[0057]

[0058] However, the mapping function f n All are unknown and cannot be directly passed through A n The key points in the ultrasound and electromagnetic data cannot cover the entire set A, and there are some mismatches, so it is not possible to directly calculate the average by averaging the key points. However, you can select a data set A a As a basis, other data sets and the data set A are obtained through feature recognition and key point matching methods. a The mapping relationship is obtained by interpolation to obtain the complete function, and the mapping function is defined so that:

[0059]

[0060] Then the above average value is:

[0061]

[0062] Step F, correcting the wellbore imaging map according to the mapping function to obtain a corrected wellbore imaging map;

[0063] In this way, we can get the correspondence between any element in set A and the element in the optimal estimation set. a For any element x in , its optimal estimated position is:

[0064]

[0065] The data corresponding to the midpoint (x, y) of the corrected image is:

[0066]

[0067]

[0068] Where: A is the location set; f a,n A a To A n The mapping function of .

[0069] The above description is not intended to limit the present invention in any form. Although the present invention has been disclosed through the above embodiments, it is not intended to limit the present invention. Any technician familiar with the profession can make some changes or modifications to equivalent embodiments of equivalent changes using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are within the scope of the technical solution of the present invention.

Claims

1. A method for locating a wellbore defect, characterized in that: The following steps are involved: Step A, synchronously timing the defect detection device, the inertial navigation system, and the three-axis magnetometer; placing the defect detection device into the wellbore, performing defect detection on the wellbore, acquiring defect detection data, inertial navigation data, and three-axis magnetometer data, and aligning them according to time; Step B, correcting the gyroscope data based on the local geomagnetic field, the trajectory data of the wellbore, the three-axis magnetometer data, and the shielding coefficient data; Step C, combining the gyroscope and accelerometer information in the inertial navigation system with the data of the scanning device to establish the well perimeter ultrasonic or magnetic flux leakage imaging, and extracting the feature points in the well perimeter imaging to obtain a feature point set; Step D: Repeat steps AC several times to obtain a data set of n measurements. and the location set , feature point set; Step E: Select any location set As the basis, the mapping function between other position sets and the basis is obtained through segmentation point selector and key point matching based on improved LightGlue algorithm; a. Feature points are segmented through segmentation point selector; b. Input the segmented feature points into the LightGlue network improved by hybrid position encoding, output the matching points, and then use the algorithm to fit them to get the fitting curve; The attention score calculation formula of the LightGlue network improved by hybrid position encoding is: Where: Q , K , V It is x query matrix, key matrix and value matrix of coordinate absolute encoding; , is the query matrix and key matrix encoded by y-coordinate rotation; c. Draw the total distortion curve according to the fitting curve, then find the inverse function and sub-pixel interpolation to obtain the mapping function; The relationship of the mapping function is: Where: is the location set; For any set of positions To location set The mapping function of Step F: Correct the wellbore imaging map according to the inverse function of the mapping function to obtain a corrected wellbore imaging map.

2. A method for locating a wellbore defect according to claim 1, characterized in that: The defect detection device is an ultrasonic or magnetic flux leakage detection device.

3. A method for locating a wellbore defect according to claim 1, characterized in that: The feature points in step C are extracted using the SIFT algorithm.

4. A method for locating a wellbore defect according to claim 3, characterized in that: The number of directional gradients of the SIFT algorithm is 8, and the sub-region grid is 4×4.

5. A method for locating a wellbore defect according to claim 1, characterized in that: The corrected formula in step F is: 。

Images