Derrick steel structure stress detection method based on vision-electromagnetic detection

Abstract

The invention discloses a derrick steel structure stress detection method based on vision-electromagnetic detection, and belongs to the field of petroleum well safety monitoring, and the method comprises the following steps: collecting a derrick image through binocular vision, constructing a current three-dimensional model and a pre-deformation three-dimensional model of a derrick, and constructing a three-dimensional model of the derrick; the derrick deformation position, the deformation degree and the derrick stress concentration position are determined according to the current three-dimensional model and the three-dimensional model before deformation, electromagnetic data of the derrick stress concentration position are collected, the derrick stress distribution condition is determined according to the electromagnetic data, a stress image is constructed, and more reliable data are provided for derrick use safety and maintenance.

Description

Technical Field

[0001] The present invention relates to the field of mechanical damage detection, and in particular to a method for detecting stress of a derrick steel structure based on visual-electromagnetic detection. Background Art

[0002] Residual stress testing is an important foundation for workpiece condition assessment. Residual stress in a workpiece can affect its hardness, fatigue state, and service life, and can even cause brittle fracture and stress corrosion cracking during use. Therefore, effective residual stress testing and assessment is a crucial means of ensuring stable and reliable workpiece performance.

[0003] Residual stress testing technology has evolved into dozens of measurement methods, primarily categorized as destructive and nondestructive. Destructive testing offers high accuracy but can damage the workpiece. Nondestructive testing primarily includes X-ray diffraction, neutron diffraction, magnetics, ultrasonics, and electronic speckle pattern interferometry. Furthermore, a derrick is a large device used during drilling or workover operations to house overhead cranes, suspend traveling blocks, hooks, eyebolts, elevators, and other equipment, as well as to raise, lower, and store drill pipe, tubing, and sucker rods. Therefore, residual stress testing on a derrick is complex and inconvenient, and a single stress testing method cannot adequately meet these requirements. Summary of the Invention

[0004] In response to the above problems, the present invention provides a derrick steel structure stress detection method based on visual-electromagnetic detection, which can facilitate the detection of derrick stress, simplify the operation, and effectively detect the stress distribution of the derrick.

[0005] The technical solution of the present invention is:

[0006] A method for detecting stress of a derrick steel structure based on visual-electromagnetic detection comprises the following steps:

[0007] S1, collecting derrick images, constructing the current 3D model of the derrick and the 3D model before deformation;

[0008] S2. Determine the deformation position and degree of the derrick based on the current three-dimensional model and the three-dimensional model before deformation, and determine the stress concentration position of the derrick;

[0009] S3. Collect electromagnetic data at the stress concentration location of the derrick, determine the stress distribution of the derrick based on the electromagnetic data, and construct a stress image.

[0010] In step S1, the derrick image is collected using a binocular vision detection system, and the current three-dimensional model is established as follows:

[0011] The left camera coordinate system is used as the world coordinate system, and the mapping point of the left camera optical center on the image plane is (x0, y0). Assuming that the parameters of the left and right cameras are the same and the space point P(X W , Y W , Z W ) The corresponding mapping point in the left and right graphs is P l (x1,y1),P r (x r ,y r ), where y1 = y r , given the camera focal length f and the binocular camera baseline length B, with the optical center of the left camera as the origin of the world coordinate system, we can obtain the following relationship:

[0012]

[0013] The compensated transformation matrix Q is as follows, where x′0 is the horizontal coordinate value of the right camera optical center mapped on the image:

[0014]

[0015] The coordinates of each pixel in the real space are calculated through the dense disparity map obtained by binocular stereo matching, and these real space points are smoothly connected to form a three-dimensional reconstruction model.

[0016] In step S1, the method for establishing the three-dimensional model before deformation is as follows:

[0017] The model is divided into n parts, F0 represents the root filter, F i (i>0) represents the i-th component filter, v i Indicates F i The horizontal and vertical displacement of the ideal position (the position of the component when no deformation occurs) relative to the position of F0, (dx i ,dy i ) represents the deviation from the ideal position of the i-th component, represents the eigenvector, φ d (dx i ,dy i ) represents the deformation characteristics, O DPM (x0, y0, z0) represents the score of the window at position (x0, y0, z0), d i Represents a four-dimensional vector parameter of a quadratic function, b is a real number representing the deviation, and the following relationship exists:

[0018]

[0019] (dx i ,dy i )=(x i ,yi )-(2(x0,y0)+v i )

[0020] φ d (dx i ,dy i )=(dx i ,dy i ,dx i 2 ,dy i 2 )

[0021] Then for n components, their offsets (dx i ,dy i ) and deformation characteristics φ d (dx i ,dy i ) are:

[0022]

[0023] The three-dimensional model before deformation is constructed by obtaining the offset and deformation characteristics of each component.

[0024] The current 3D model and the 3D model before deformation need to be corrected after being established. The correction method is as follows:

[0025] The images captured by the binocular camera are calibrated. The parameters obtained by the binocular camera calibration include the rotation matrix R and translation vector T of the right camera relative to the left camera, and these two parameters are needed in the correction algorithm. The rotation matrix R is decomposed into two matrices R l and R r , the left and right cameras are aligned according to the matrix R l and R r After the rotation, the image planes of the two cameras are made coplanar. Apply the transformation matrix R rect Align the rows of the left and right images, transform matrix R rect is constructed as follows:

[0026]

[0027] T=[T x T y T z ]

[0028]

[0029] Vector e2 is orthogonal to vector e1 and has the same direction as the image plane. Vector e3 is perpendicular to the plane containing vectors e2 and e1. The above algorithm is applied to perform stereo correction on the camera to eliminate the distortion introduced by the camera lens and improve the geometric accuracy and quality of the image.

[0030] In step S3, when collecting electromagnetic data, a full-cycle digital lock-in amplification algorithm is used to extract weak signals;

[0031] Only signals sampled at full cycles will not experience spectrum leakage. For a signal to be measured with a period of T, it is sampled at intervals of τ to obtain a signal sequence. The length of the signal sequence is M. If M·τ=q·T, and q is an integer, then the signal is said to have achieved full cycle sampling. By sampling the signal in full cycle sampling to obtain a signal sequence, digital operations are used to accurately obtain a reference sequence with the same frequency. By calculating the digital cross-correlation between the signal sequence and the reference sequence, the feature extraction of the signal to be measured can be achieved. The specific algorithm is as follows:

[0032] Assume that the frequency of the signal to be measured is f r , the sampling frequency is f s , control the sampling frequency so that f s =M·f r / q, A is the signal amplitude, is the initial phase of the signal, and M samples are taken in q reference signal cycles, with a sampling interval of τ = 1 / f s =q / (M·f r ), then the signal sequence x(k):

[0033]

[0034] When sampling in the whole cycle, the sine reference sequence rs(k) and the cosine reference sequence rc(k) are directly obtained by digital operation, and then the cross-correlation R between the signal sequence and the sine reference sequence is calculated respectively. xrs and the cross-correlation R with the cosine reference sequence xrc , and finally calculate the amplitude A and phase of the signal Thus, the feature extraction of the signal to be measured is completed;

[0035]

[0036] Where rs(k) represents the sine reference sequence, rc(k) represents the cosine reference sequence, q represents the signal period, M represents the number of sampling times, and R xrs represents the cross-correlation between the signal sequence and the sinusoidal reference sequence, R xrc The amplitude A and phase of the signal calculated by the above formula are The signal sequence x(k) can be finally calculated, thereby realizing the feature extraction of the signal to be measured.

[0037] In step S3, the stress distribution is determined as follows:

[0038] The measured position of the derrick is divided into N metal thin layers with a thickness of d and a length of l. The conductivity of each layer is σ1, σ2, ..., σ N , using frequencies f, f / 2 respectively 2 ,...,f / N 2 The excitation signal is used to perform frequency sweep detection. For the kth excitation signal, the operating frequency is f / k 2 , assuming that the effective cross-sectional area of ​​the top metal layer is S, the effective cross-sectional area of ​​the i-th layer is approximately e represents a natural constant, then the total conductance Y of the N layer is k for:

[0039]

[0040] Introducing the parameter q, q = l / S, where l is the length of the metal layer and S is the effective cross-sectional area of ​​the metal layer. Writing the above formula in matrix form, we can get the inversion model:

[0041]

[0042] The above inversion model can be used to calculate the conductivity σ1, σ2, ..., σ at different depths of the derrick. N , based on the stress and conductivity model of the empirical formula, the relationship between stress and conductivity is obtained as follows:

[0043]

[0044] where δ i represents the stress of the i-th layer, C is a coefficient that depends on the specific material, μ represents Poisson's ratio, E represents Young's modulus, and σ0 represents the initial conductivity. According to the relationship between stress and conductivity, the stresses δ1, δ2, ..., δ at different depths of the measured position of the derrick can be calculated. N , thus obtaining the stress distribution at the measured position of the derrick.

[0045] The beneficial effects of the present invention are:

[0046] 1. Use binocular vision inspection technology to first inspect the entire derrick and preliminarily determine the location of stress concentration, reducing workload. Instead of performing precise stress testing on every location of the derrick, the entire derrick can be inspected, shortening inspection time and improving inspection efficiency.

[0047] 2. Based on the preliminary judgment of the stress concentration location, specific stress detection is carried out on the location where the stress is concentrated on the derrick through electromagnetic detection technology, which can accurately obtain the stress distribution of the derrick, facilitate the prediction of the service life of the derrick, and replace the derrick in time. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a method flow chart of a method for detecting stress in a derrick steel structure based on visual-electromagnetic detection according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0050] Example:

[0051] like Figure 1 As shown, a method for detecting stress of a derrick steel structure based on visual-electromagnetic detection includes the following steps.

[0052] S1. The image acquisition system uses a binocular camera to capture the plan view of the derrick at different positions and stores it in the camera.

[0053] S2. Import the derrick image data stored in S1 into the data processing module to construct a current three-dimensional stereogram of the derrick and a three-dimensional stereogram before deformation.

[0054] Method for constructing the current three-dimensional stereogram:

[0055] The left camera coordinate system is used as the world coordinate system, and the mapping point of the left camera optical center on the image plane is (x0, y0). Assuming that the parameters of the left and right cameras are the same and the space point P(X W , Y W , Z W ) The corresponding mapping point in the left and right graphs is P l (x1,y1),P r (x r ,y r ), where y1 = y r , given the camera focal length f and the binocular camera baseline length B, with the optical center of the left camera as the origin of the world coordinate system, we can obtain the following relationship:

[0056]

[0057] The compensated transformation matrix Q is as follows, where x′0 is the horizontal coordinate value of the right camera optical center mapped on the image:

[0058]

[0059] The coordinates of each pixel in the real space are calculated through the dense disparity map obtained by binocular stereo matching, and these real space points are smoothly connected to form a three-dimensional reconstruction model.

[0060] Method for constructing a 3D stereogram before deformation:

[0061] The model is divided into n parts, F0 represents the root filter, F i (i>0) represents the i-th component filter, v i Indicates F i The horizontal and vertical displacement of the ideal position (the position of the component when no deformation occurs) relative to the position of F0, (dx i ,dy i ) represents the deviation from the ideal position of the i-th component, represents the eigenvector, φ d (dx i ,dy i ) represents the deformation characteristics, O DPM (x0, y0, z0) represents the score of the window at position (x0, y0, z0), d i Represents a four-dimensional vector parameter of a quadratic function, b is a real number representing the deviation, and the following relationship exists:

[0062]

[0063] (dx i ,dy i )=(x i ,y i )-(2(x0,y0)+v i )

[0064] φ d (dx i ,dy i )=(dx i ,dy i ,dx i 2 ,dy i 2 )

[0065] Then for n components, their offsets (dx i ,dy i ) and deformation characteristics φ d (dx i ,dy i ) are:

[0066]

[0067] By obtaining the offset and deformation characteristics of each component, these data information are integrated to construct a three-dimensional model before deformation.

[0068] S3. Import the three-dimensional model in S2 into the deformation calculation module to determine the deformation position and deformation degree of the derrick and the location where the stress of the derrick is concentrated.

[0069] S4. Performing specific stress detection on the location where stress concentration occurs on the derrick through an electromagnetic detection system, wherein the electromagnetic detection system includes an electromagnetic signal acquisition module, an electromagnetic signal detection module, and an upper computer information processing module;

[0070] The electromagnetic signal acquisition module includes an electromagnetic detection sensor and a preamplifier board; the electromagnetic signal detection module includes a control communication module with FPGA as the core, the FPGA module includes a high-speed AD acquisition circuit, a DDS multi-frequency excitation signal generation circuit, and a DDR memory read-write controller. The electromagnetic signal detection module also includes a signal conditioning circuit and a DSP digital signal processor; the upper computer information processing module includes parameter and control command setting, data calibration, temperature calibration, and stress detection and imaging.

[0071] S5. The electromagnetic detection sensor in the electromagnetic signal acquisition module will output the signal to be tested under sinusoidal excitation. The signal to be tested enters the preamplifier board for conditioning and amplification, and then is sent to the electromagnetic signal detection module for signal feature extraction. The high-speed AD acquisition circuit collects the signal, and then the collected data is written into the DDR through the memory, and then enters the DSP digital signal processor to extract and process the information.

[0072] Feature extraction method of the signal to be measured:

[0073] Only signals sampled in full cycles will not have spectrum leakage. For a signal to be measured with a period of T, it is sampled at intervals of τ to obtain a signal sequence. The length of the signal sequence is M. If M·τ=q·T, and q is an integer, then the signal is said to have achieved full-cycle sampling. When collecting electromagnetic data, the full-cycle digital lock-in amplification algorithm is used to extract weak signals. During full-cycle sampling, the signal is sampled to obtain a signal sequence. A reference sequence with the same frequency is accurately obtained through digital calculations. Then, the digital cross-correlation between the signal sequence and the reference sequence is calculated to achieve feature extraction of the signal to be measured. The specific algorithm is as follows:

[0074] Assume that the frequency of the signal to be measured is f r , the sampling frequency is f s , control the sampling frequency so that f s =M·f r / q, A is the signal amplitude, is the initial phase of the signal, and M samples are taken in q reference signal cycles, with a sampling interval of τ = 1 / f s=q / (M·f r ), then the signal sequence x(k):

[0075]

[0076] When sampling in the whole cycle, the sine reference sequence rs(k) and the cosine reference sequence rc(k) are directly obtained according to digital operations, and then the cross-correlation R between the signal sequence and the sine reference sequence is calculated respectively. xrs and the cross-correlation R with the cosine reference sequence xrc , and finally calculate the amplitude A and phase of the signal Thus, the feature extraction of the signal to be measured is completed;

[0077]

[0078] Where rs(k) represents the sine reference sequence, rc(k) represents the cosine reference sequence, q represents the signal period, M represents the number of sampling times, and R xrs represents the cross-correlation between the signal sequence and the sinusoidal reference sequence, R xrc The amplitude A and phase of the signal calculated by the above formula are The signal sequence x(k) can be finally calculated, thereby realizing the feature extraction of the signal to be measured.

[0079] S6. Finally, the host computer interacts with the DSP digital signal processor, and the host computer information processing module performs sensor calibration, stress distribution detection and stress imaging on the received data.

[0080] Method for detecting stress distribution by host computer:

[0081] The measured position of the derrick is divided into N metal thin layers with a thickness of d and a length of l. The conductivity of each layer is σ1, σ2, ..., σ N , using frequencies f, f / 2 respectively 2 ,...,f / N 2 The excitation signal is used to perform frequency sweep detection. For the kth excitation signal, the operating frequency is f / k 2 , assuming that the effective cross-sectional area of ​​the top metal layer is S, the effective cross-sectional area of ​​the i-th layer is approximately e represents a natural constant, then the total conductance Y of the N layer is k for:

[0082]

[0083] The parameter q is introduced to make q = l / S. The parameter q itself has no meaning, but after introducing the parameter q, the conductivity at different depths of the measured position of the derrick can be calculated more easily, where l is the length of the metal layer and S is the effective cross-sectional area of ​​the metal layer. The inversion model can be obtained by writing the above formula in matrix form:

[0084]

[0085] The above inversion model can be used to calculate the conductivity σ1, σ2, ..., σ at different depths of the derrick. N , based on the stress and conductivity model of the empirical formula, the relationship between stress and conductivity is obtained as follows:

[0086]

[0087] where δ i represents the stress of the i-th layer, C is a coefficient that depends on the specific material, μ represents Poisson's ratio, E represents Young's modulus, and σ0 represents the initial conductivity. According to the relationship between stress and conductivity, the stresses δ1, δ2, ..., δ at different depths of the measured position of the derrick can be calculated. N , thus obtaining the stress distribution at the measured position of the derrick.

[0088] The above-described embodiments merely represent specific implementations of the present invention. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, and all such variations and improvements fall within the scope of protection of the present invention.

Claims

1. A method for detecting stress of derrick steel structure based on visual-electromagnetic detection, characterized in that: The detection method includes the following steps: S1. Collecting images of the derrick and constructing a current 3D model of the derrick and a 3D model before deformation; S2. Determine the deformation position and degree of the derrick based on the current three-dimensional model and the three-dimensional model before deformation, and determine the stress concentration position of the derrick; S3. Collect electromagnetic data at the stress concentration location of the derrick, determine the stress distribution of the derrick based on the electromagnetic data, and construct a stress image.

2. The method for detecting stress of a derrick steel structure based on visual-electromagnetic detection according to claim 1, characterized in that: In step S1, the derrick image is collected using a binocular vision detection system, and the current three-dimensional model is established as follows: The left camera coordinate system is used as the world coordinate system, and the mapping point of the left camera optical center on the image plane is (x0, y0). Assuming that the parameters of the left and right cameras are the same and the space point P(X W , Y W , Z W ) The corresponding mapping point in the left and right graphs is P l (x1,y1),P r (x r ,y r ), where y1 = y r , given the camera focal length f and the binocular camera baseline length B, with the optical center of the left camera as the origin of the world coordinate system, we can obtain the following relationship: The compensated transformation matrix Q is as follows, where x′0 is the horizontal coordinate value of the right camera optical center mapped on the image: The coordinates of each pixel in the real space are calculated through the dense disparity map obtained by binocular stereo matching, and these real space points are smoothly connected to form a three-dimensional reconstruction model.

3. The method for detecting stress of a derrick steel structure based on visual-electromagnetic detection according to claim 1, characterized in that: In step S1, the method for establishing the three-dimensional model before deformation is as follows: The model is divided into n parts, F0 represents the root filter, F i (i>0) represents the i-th component filter, v i Indicates F i The horizontal and vertical displacement of the ideal position (the position of the component when no deformation occurs) relative to the position of F0, (dx i ,dy i ) represents the deviation from the ideal position of the i-th component, represents the eigenvector, φ d (dx i ,dy i ) represents the deformation characteristics, O DPM (x0, y0, z0) represents the score of the window at position (x0, y0, z0), d i Represents a four-dimensional vector parameter of a quadratic function, b is a real number representing the deviation, and the following relationship exists: (dx i ,dy i )=(x i ,y i )-(2(x0,y0)+v i ) φ d (dx i ,of i )=(dx i ,of i ,dx i 2 ,of i 2 ) Then for n components, their offsets (dx i ,dy i ) and deformation characteristics φ d (dx i ,dy i ) are: The three-dimensional model before deformation is constructed by obtaining the offset and deformation characteristics of each component.

4. A method for detecting stress of a derrick steel structure based on visual-electromagnetic detection according to claim 2 or 3, characterized in that: The current 3D model and the 3D model before deformation need to be corrected after being established. The correction method is as follows: The images captured by the binocular camera are calibrated. The parameters obtained by the binocular camera calibration include the rotation matrix R and translation vector T of the right camera relative to the left camera, and these two parameters are needed in the correction algorithm. The rotation matrix R is decomposed into two matrices R l and R r , the left and right cameras are aligned according to the matrix R l and R r After the rotation, the image planes of the two cameras are made coplanar. Apply the transformation matrix R rect Align the rows of the left and right images, transform matrix R rect is constructed as follows: T=[T x T y T z ] Vector e2 is orthogonal to vector e1 and has the same direction as the image plane. Vector e3 is perpendicular to the plane containing vectors e2 and e1. The above algorithm is applied to perform stereo correction on the camera to eliminate the distortion introduced by the camera lens and improve the geometric accuracy and quality of the image.

5. The method for detecting stress of a derrick steel structure based on visual-electromagnetic detection according to claim 1, characterized in that: In step S3, when collecting electromagnetic data, a full-cycle digital lock-in amplification algorithm is used to extract weak signals; The signal is sampled in the full cycle sampling to obtain a signal sequence. The reference sequence with the same frequency is accurately obtained through digital calculation. Then the digital cross-correlation between the signal sequence and the reference sequence is calculated to realize the feature extraction of the signal to be measured. The specific algorithm is as follows: Assume that the frequency of the signal to be measured is f r , the sampling frequency is f s , control the sampling frequency so that f s =M·f r / q, A is the signal amplitude, is the initial phase of the signal, and M samples are taken in q reference signal cycles, with a sampling interval of τ = 1 / f s =q / (M·f r ), then the signal sequence x(k): When sampling in the whole cycle, the sine reference sequence rs(k) and the cosine reference sequence rc(k) are directly obtained according to digital operations, and then the cross-correlation R between the signal sequence and the sine reference sequence is calculated respectively. xrs and the cross-correlation R with the cosine reference sequence xrc , and finally calculate the amplitude A and phase of the signal Thus, the feature extraction of the signal to be measured is completed; Where rs(k) represents the sine reference sequence, rc(k) represents the cosine reference sequence, q represents the signal period, M represents the number of sampling times, and R xrs represents the cross-correlation between the signal sequence and the sinusoidal reference sequence, R xrc The amplitude A and phase of the signal calculated by the above formula are The signal sequence x(k) can be finally calculated, thereby realizing the feature extraction of the signal to be measured.

6. The method for detecting stress of a derrick steel structure based on visual-electromagnetic detection according to claim 1, characterized in that: In step S3, the stress distribution is determined as follows: The measured position of the derrick is divided into N metal thin layers with a thickness of d and a length of l. The conductivity of each layer is σ1, σ2, ..., σ N , using frequencies f, f / 2 respectively 2 ,...,f / N 2 The excitation signal is used to perform frequency sweep detection. For the kth excitation signal, the operating frequency is f / k 2 , assuming that the effective cross-sectional area of ​​the top metal layer is S, the effective cross-sectional area of ​​the i-th layer is approximately e represents a natural constant, then the total conductance Y of the N layer is k for: Introducing the parameter q, q = l / S, where l is the length of the metal layer and S is the effective cross-sectional area of ​​the metal layer. Writing the above formula in matrix form, we can get the inversion model: The above inversion model can be used to calculate the conductivity σ1, σ2, ..., σ at different depths of the derrick. N , based on the stress and conductivity model of the empirical formula, the relationship between stress and conductivity is obtained as follows: where δ i represents the stress of the i-th layer, C is a coefficient that depends on the specific material, μ represents Poisson's ratio, E represents Young's modulus, and σ0 represents the initial conductivity. According to the relationship between stress and conductivity, the stresses δ1, δ2, ..., δ at different depths of the measured position of the derrick can be calculated. N , thus obtaining the stress distribution at the measured position of the derrick.

Images