Non-contact rotating body vibration detection method and system

Abstract

The invention discloses a non-contact rotating body vibration detection method and system, and the method comprises the steps: obtaining an internal parameter matrix and an external parameter matrix of a high-speed camera through employing a Zhang's calibration method, and completing the conversion between a world coordinate system and an image coordinate system, thereby completing the initial error correction of the high-speed camera; then adding feature points to the rotating body and obtaining the rotating body after the feature points are added, and shooting the rotating body after the feature points are added by using the corrected high-speed camera so as to obtain a rotating body image with the feature points; according to the invention, a function of carrying out high-real-time reliable high-precision vibration detection on the rotating body in a non-contact detection mode is realized, the limitation of an existing contact detection method is overcome, high-precision, real-time and reliable vibration characteristic detection is realized, accidental damage of equipment is avoided, the downtime is reduced, and the working efficiency is improved. The production safety and the production efficiency of the equipment are ensured, and the device is suitable for wide popularization and use.

Description

Technical Field

[0001] The present invention relates to the technical field of rotating body vibration detection, and particularly relates to a non-contact rotating body vibration detection method and system. Background Art

[0002] Rotating equipment refers to mechanical equipment whose main function is completed by rotational motion, including steam turbines, gas turbines, centrifugal compressors, generators, water pumps, fans, and motors.

[0003] Currently, rotating equipment will inevitably be subjected to different types of vibrations during operation. If these vibrations exceed the normal range, it may indicate wear, imbalance, or potential faults in the equipment. Most of the existing vibration detection methods use contact detection methods for detection, resulting in large deviations in vibration detection results. This not only easily damages the equipment, but also increases the downtime, reduces the production efficiency, and poses serious safety hazards. Therefore, it is necessary to design a non-contact rotating body vibration detection method and system. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies of the prior art. In order to better and effectively solve the problem that the existing vibration detection methods mostly use contact detection methods for detection, resulting in large deviations in vibration detection results, which not only easily damage the equipment, but also increase the downtime, reduce the production efficiency, and pose serious safety hazards, a non-contact rotating body vibration detection method and system are provided. It realizes the function of high-real-time, reliable, and high-precision vibration detection of the rotating body by using a non-contact detection method. It not only overcomes the limitations of the existing contact detection methods, but also realizes high-precision, real-time, and reliable vibration feature detection, avoids accidental damage to the equipment, reduces the downtime, and ensures the safety and production efficiency of the equipment.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0006] A non-contact rotating body vibration detection method includes the following steps:

[0007] Step A: Use the Zhang calibration method to obtain the internal parameter matrix and external parameter matrix of the high-speed camera and complete the conversion between the world coordinate system and the image coordinate system, thereby completing the preliminary error correction of the high-speed camera.

[0008] Step B: Add feature points to the rotating body to obtain the rotating body with feature points added, and then use the corrected high-speed camera to photograph the rotating body with feature points added to obtain a rotating body image with feature points.

[0009] Step C: Apply the YOLOv7 algorithm to the obtained rotating body image with feature points to extract the image feature points, and obtain the target bounding box and class information;

[0010] Step D: Denoise the target bounding box and class information using the SVD denoising method to obtain the vibration displacement information, and complete the non-contact detection operation of the rotating body.

[0011] For the aforementioned non-contact rotating body vibration detection method, in Step A, the Zhang calibration method is used to obtain the internal parameter matrix and external parameter matrix of the high-speed camera and complete the conversion between the world coordinate system and the image coordinate system, thereby completing the preliminary error correction of the high-speed camera. The specific steps are as follows.

[0012] Step A1: Use the Zhang calibration method to determine the correspondence between the three-dimensional geometric position of points on an object in space and the two-dimensional geometric position of the corresponding points of these points on the object in the image, as shown in formula (1).

[0013]

[0014] where, [X w Y w Z w 1] T and [X c Y c Z c 1] T are the homogeneous coordinates of any point in space in the world coordinate system and the camera coordinate system respectively, (R, t) are the external parameters of the camera, where R is the rotation matrix and t is the translation vector;

[0015] Step A2: Represent the relationship between the image coordinate system and the world coordinate system using the geometric imaging principle in the Zhang calibration method, as shown in formula (2).

[0016]

[0017] where, and are the homogeneous coordinates of point M and point m respectively, s is the scale factor, and γ is the skew coefficient of the image coordinate system;

[0018] Step A3: When the world coordinate system and the camera coordinate system coincide, and at this time the world coordinate system plane is placed on the calibration plate plane, rewrite formula (2) as:

[0019]

[0020] where the matrix A[r 1 r 2 t] is the mixed matrix of the internal parameter matrix and the external parameter matrix.

[0021] The above non-contact rotating body vibration detection method, step B: Add feature points to the rotating body and obtain the rotating body after adding the feature points. Then, use the calibrated high-speed camera to photograph the rotating body after adding the feature points to obtain the rotating body image with feature points. If the rotating body has natural feature points, there is no need to artificially add feature points during shooting. If the rotating body has no natural feature points, different color points are used to increase the feature points of the rotating body. The specific steps of using the high-speed camera to photograph the rotating body with feature points are as follows:

[0022] Step B1: Use the illumination light source to illuminate the shooting environment, then use the optical lens to image the rotating body on the image sensor, and then convert the photographed image into an analog signal;

[0023] Step B2: Use the image acquisition card to convert the analog signal into a digital image signal, and then transmit the digital image signal to the computer processing system.

[0024] The above non-contact rotating body vibration detection method, step C: Use the YOLOv7 algorithm to extract the image feature points of the obtained rotating body image with feature points to obtain the target bounding box and class information. The specific steps are as follows:

[0025] Step C1: Grid division. Specifically, divide the obtained rotating body image with feature points into grids of a fixed size, and each grid is responsible for detecting the targets inside it and predicting the bounding box;

[0026] Step C2: Bounding box prediction. The bounding box inside each grid includes position information and class information, and the bounding box prediction Confidence is shown in formula (4):

[0027] Confidence = P(object) × IOU(4)

[0028] Where, P(object) is the target existence probability, and IOU is the intersection over union of the predicted box and the ground truth box;

[0029] Step C3: Target score calculation. Specifically, calculate the target score of each bounding box, and the target score of each bounding box indicates whether the target is included inside the bounding box;

[0030] Step C4: Non-maximum suppression. Specifically, eliminate the overlapping bounding boxes and only retain the bounding box with the highest score, and output the target bounding box and class information.

[0031] The above non-contact rotating body vibration detection method, step D: Use the SVD denoising method to denoise the target bounding box and class information to obtain the vibration displacement information. The specific steps are as follows:

[0032] Step D1: Extract the displacement signals in the x - direction and y - direction from the image containing the target bounding box and class information, and construct them into a binary recursive matrix A respectively i-1 , as shown in formula (5):

[0033]

[0034] Step D2: Perform SVD processing on formula (5), and the processing result is shown in formula (6):

[0035] A i = U i D i V i T (6)

[0036] where U i (u i1 , u i2 ) is the left singular matrix obtained from the i - th decomposition, V i (v i1 , v i2 ) is the right singular matrix obtained from the i - th decomposition, and D i is a diagonal matrix;

[0037] After the i - th SVD processing, it is as shown in formula (7):

[0038] A i = σ ai u i1 v i1 T + σ bi u i2 v i2 T (7)

[0039] where v i1 and v i2 are the sub - elements of V i respectively, σ ai u i1 v i1 T and σ bi u i2 v i2 T are the first - order singular value and the second - order singular value respectively;

[0040] Step D4: Sum and take the average of all elements in the two column vectors of the matrix σ ai u i1 v i1 T , as shown in formula (8):

[0041]

[0042] A non-contact rotating body vibration detection system includes an error correction module, an image acquisition module, a feature extraction module, and a noise processing module. The error correction module is used to obtain the internal parameter matrix and external parameter matrix of the high-speed camera using the Zhang's calibration method and complete the conversion between the world coordinate system and the image coordinate system, thereby completing the preliminary error correction of the high-speed camera. The image acquisition module is used to add feature points to the rotating body and obtain the rotating body with the added feature points, and then use the corrected high-speed camera to capture the rotating body with the added feature points to obtain a rotating body image with feature points. The feature extraction module is used to extract the image feature points of the obtained rotating body image with feature points using the YOLOv7 algorithm and obtain the target bounding box and class information. The noise processing module is used to denoise the target bounding box and class information using the SVD denoising method and obtain the vibration displacement information, thereby completing the non-contact detection operation of the rotating body.

[0043] The beneficial effects of the present invention are as follows: For a non-contact rotating body vibration detection method and system of the present invention, first, the internal parameter matrix and external parameter matrix of the high-speed camera are obtained using the Zhang's calibration method and the conversion between the world coordinate system and the image coordinate system is completed, thereby completing the preliminary error correction of the high-speed camera. Then, feature points are added to the rotating body and the rotating body with the added feature points is obtained, and then the corrected high-speed camera is used to capture the rotating body with the added feature points, thereby obtaining a rotating body image with feature points. Subsequently, the image feature points of the obtained rotating body image with feature points are extracted using the YOLOv7 algorithm and the target bounding box and class information are obtained. Then, the target bounding box and class information are denoised using the SVD denoising method and the vibration displacement information is obtained, thereby completing the non-contact detection operation of the rotating body. It effectively realizes that the non-contact rotating body vibration detection method and system have the function of performing high-real-time, reliable, and high-precision vibration detection on the rotating body using a non-contact detection method. It not only overcomes the limitations of the existing contact detection methods, but also realizes high-precision, real-time, and reliable vibration feature detection, avoids accidental damage to the equipment, reduces the downtime, ensures the safety of equipment production and the production efficiency of the equipment. At the same time, it can adapt to complex operating environments and also reduces the interference to the installation and operation of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is the overall flowchart of a non-contact rotating body vibration detection method of the present invention;

[0045] Figure 2 is the flowchart of rotating body image acquisition of the present invention;

[0046] Figure 3It is the signal acquisition and detection flow chart of the present invention;

[0047] Figure 4 It is the amplitude schematic diagram of the detection result of the present invention;

[0048] Figure 5 It is the frequency schematic diagram of the detection result of the present invention. Specific embodiments

[0049] Next, the present invention will be further described in conjunction with the accompanying drawings of the specification.

[0050] As Figures 1-5 shown, a non-contact rotating body vibration detection method of the present invention includes the following steps.

[0051] Step A, using Zhang's calibration method to obtain the internal parameter matrix and external parameter matrix of the high-speed camera and complete the conversion between the world coordinate system and the image coordinate system, so as to complete the preliminary error correction of the high-speed camera. The specific steps are as follows.

[0052] Step A1, using Zhang's calibration method to determine the corresponding relationship between the three-dimensional geometric position of a point on an object in space and the two-dimensional geometric position of the corresponding point of the point on the object in the image, as shown in formula (1).

[0053]

[0054] Among them, [X w Y w Z w 1] T and [X c Y c Z c 1] T are the homogeneous coordinates of any point in space in the world coordinate system and the camera coordinate system respectively. (R, t) are the external parameters of the camera, where R is the rotation matrix and t is the translation vector.

[0055] Step A2, using the geometric imaging principle to represent the relationship between the image coordinate system and the world coordinate system in Zhang's calibration method, as shown in formula (2).

[0056]

[0057] Among them, and are the homogeneous coordinates of point M and point m respectively. s is the scale factor and γ is the skew coefficient of the image coordinate system.

[0058] Step A3, when the world coordinate system and the camera coordinate system coincide, at this time the world coordinate system plane is placed on the calibration plate plane, rewrite formula (2) as:

[0059]

[0060] Among them, the matrix A[r 1 r 2 t] is a mixed matrix of the internal parameter matrix and the external parameter matrix.

[0061] Step B: Add feature points to the rotating body to obtain the rotating body with added feature points, and then use the calibrated high-speed camera to photograph the rotating body with added feature points to obtain the rotating body image with feature points. If the rotating body has natural feature points, there is no need to artificially add feature points during shooting. If the rotating body has no natural feature points, different colored points are used to increase the feature points of the rotating body. The specific steps of using the high-speed camera to photograph the rotating body with feature points are as follows.

[0062] Step B1: Use the illumination light source to illuminate the shooting environment, then use the optical lens to image the rotating body on the image sensor, and then convert the captured image into an analog signal.

[0063] Step B2: Use the image acquisition card to convert the analog signal into a digital image signal, and then transmit the digital image signal to the computer processing system.

[0064] Step C: Use the YOLOv7 algorithm to extract the image feature points of the obtained rotating body image with feature points to obtain the target bounding box and class information. The specific steps are as follows.

[0065] Step C1: Grid division. Specifically, divide the obtained rotating body image with feature points into grids of a fixed size, and each grid is responsible for detecting the objects inside it and predicting the bounding box.

[0066] Step C2: Bounding box prediction. Among them, the bounding box in each grid includes position information and class information, and the bounding box prediction Confidence is shown in formula (4).

[0067] Confidence = P(object) × IOU(4)

[0068] Among them, P(object) is the target existence probability, and IOU is the intersection over union of the predicted box and the ground truth box.

[0069] Step C3: Target score calculation. Specifically, calculate the target score of each bounding box, and the target score of each bounding box indicates whether the target is included in the bounding box.

[0070] Step C4: Non-maximum suppression. Specifically, eliminate the overlapping bounding boxes and only retain the bounding box with the highest score, and output the target bounding box and class information.

[0071] Step D: Denoise the target bounding box and class information using the SVD denoising method to obtain vibration displacement information. The specific steps are as follows:

[0072] Step D1: Extract the displacement signals in the x and y directions from the image containing the target bounding box and class information and construct them into a binary recursive matrix A respectively, as shown in Equation (5). i-1 , as shown in Equation (5).

[0073]

[0074] Step D2: Perform SVD processing on Equation (5). The processing result is shown in Equation (6).

[0075] A i = U i D i V i T (6)

[0076] where U i (u i1 , u i2 ) is the left singular matrix obtained from the i-th decomposition, V i (v i1 , v i2 ) is the right singular matrix obtained from the i-th decomposition, and D i is a diagonal matrix;

[0077] Step D3: After the i-th SVD processing, the result is shown in Equation (7).

[0078] A i = σ ai u i1 v i1 T + σ bi u i2 v i2 T (7)

[0079] where v i1 and v i2 are the sub-elements of V i respectively, and σ ai u i1 v i1 T and σ bi u i2 v i2 T are the first-order singular value and the second-order singular value respectively;

[0080] Step D4: For the matrix σ ai u i1 v i1T Sum all the elements in the two column vectors and take the average, as shown in formula (8).

[0081]

[0082] A non-contact rotating body vibration detection system includes an error correction module, an image acquisition module, a feature extraction module, and a noise processing module. The error correction module is used to obtain the internal parameter matrix and external parameter matrix of the high-speed camera by using the Zhang's calibration method and complete the conversion between the world coordinate system and the image coordinate system, so as to complete the preliminary error correction of the high-speed camera. The image acquisition module is used to add feature points to the rotating body and obtain the rotating body with the added feature points, and then use the corrected high-speed camera to take pictures of the rotating body with the added feature points and obtain the rotating body image with feature points. The feature extraction module is used to extract the image feature points of the obtained rotating body image with feature points by using the YOLOv7 algorithm and obtain the target bounding box and class information. The noise processing module is used to denoise the target bounding box and class information by using the SVD denoising method and obtain the vibration displacement information, so as to complete the non-contact detection operation of the rotating body.

[0083] In summary, for a non-contact rotating body vibration detection method and system of the present invention, first, the internal parameter matrix and external parameter matrix of the high-speed camera are obtained by using the Zhang's calibration method and the conversion between the world coordinate system and the image coordinate system is completed, so as to complete the preliminary error correction of the high-speed camera. Then, feature points are added to the rotating body and the rotating body with the added feature points is obtained. Then, the corrected high-speed camera is used to take pictures of the rotating body with the added feature points, so as to obtain the rotating body image with feature points. Subsequently, the image feature points of the obtained rotating body image with feature points are extracted by using the YOLOv7 algorithm and the target bounding box and class information are obtained. Then, the target bounding box and class information are denoised by using the SVD denoising method and the vibration displacement information is obtained, so as to complete the non-contact detection operation of the rotating body. It effectively realizes that the non-contact rotating body vibration detection method and system have the function of high-real-time, reliable and high-precision vibration detection of the rotating body by using a non-contact detection method. It not only overcomes the limitations of the existing contact detection method, but also realizes high-precision, real-time and reliable vibration feature detection, avoids accidental damage to the equipment, reduces the downtime, ensures the equipment production safety and equipment production efficiency, can adapt to complex operating environments at the same time, and also reduces the interference to the equipment installation and operation.

[0084] The basic principles, main features and advantages of the present invention have been shown and described above. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification only illustrates the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. A non-contact rotating body vibration detection method, characterized in that: The following steps are included: Step A, using Zhang's calibration method to obtain the internal parameter matrix and the external parameter matrix of the high-speed camera and complete the conversion between the world coordinate system and the image coordinate system, thereby completing the preliminary error correction of the high-speed camera; Step B, adding feature points to the rotating body and obtaining the rotating body after the feature points are added, and then photographing the rotating body after the feature points are added using a rectified high-speed camera to obtain an image of the rotating body with the feature points; Step C, extracting image feature points from the obtained rotating body image with feature points using the YOLOv7 algorithm to obtain the target bounding box and category information; In step D, the target bounding box and category information are denoised using the SVD denoising method to obtain vibration displacement information and complete the non-contact detection of the rotating object.

2. A non-contact rotating body vibration detection method according to claim 1, characterized in that: Step A, using Zhang's calibration method to obtain the internal parameter matrix and external parameter matrix of the high-speed camera and complete the conversion between the world coordinate system and the image coordinate system, so as to complete the preliminary error correction of the high-speed camera. The specific steps are as follows: Step A1, using Zhang's calibration method to determine the correspondence between the three-dimensional geometric position of a point on the object in space and the two-dimensional geometric position of the corresponding point on the object in space in the image, as shown in formula (1), Among them, [X w Y w Z w 1] T and [X c Y c Z c 1] T are the homogeneous coordinates of any point in space in the world coordinate system and the camera coordinate system, respectively, (R, t) are the external parameters of the camera, R is the rotation matrix, and t is the translation vector; Step A2, using the geometric imaging principle, the relationship between the image coordinate system and the world coordinate system is expressed in Zhang's calibration method, as shown in formula (2): in, and are the homogeneous coordinates of point M and point m, s is the scale factor, and γ is the skew coefficient of the image coordinate system; Step A3, when the world coordinate system and the camera coordinate system coincide, the world coordinate system plane is placed on the plane where the calibration plate is located, and equation (2) is rewritten as: Among them, the matrix A[r1 r2 t] is a mixed matrix of the internal parameter matrix and the external parameter matrix.

3. A non-contact rotating body vibration detection method according to claim 2, characterized in that: Step B, adding feature points to the rotating body and obtaining the rotating body after adding the feature points, and then using the corrected high-speed camera to shoot the rotating body after adding the feature points to obtain the rotating body image with feature points, wherein if the rotating body has natural feature points, there is no need to add feature points artificially during shooting, if the rotating body has no natural feature points, then different color points are used to increase the feature points of the rotating body, and the specific steps of using the high-speed camera to shoot the rotating body with feature points are as follows, Step B1, using an illumination light source to illuminate the shooting environment, then using an optical lens to image the rotating body on an image sensor, and then converting the captured image into an analog signal; Step B2, using an image acquisition card to convert the analog signal into a digital image signal, and then transmitting the digital image signal to a computer processing system.

4. A non-contact rotating body vibration detection method according to claim 3, characterized in that: Step C: extract the image feature points of the obtained rotating body image with feature points using the YOLOv7 algorithm to obtain the target bounding box and category information. The specific steps are as follows: Step C1, grid division, specifically, dividing the obtained rotating body image with feature points into grids of fixed size, and each grid is responsible for detecting its internal target and predicting the bounding box; Step C2, bounding box prediction, where the bounding box in each grid includes location information and category information, and the bounding box prediction Confidence is shown in formula (4): Confidence=P(object)×IOU(4) Among them, P(object) is the probability of the target existence, and IOU is the intersection-union ratio between the predicted box and the real box; Step C3, target score calculation, specifically, calculating the target score of each bounding box, and the target score of each bounding box indicates whether the bounding box contains the target; Step C4, non-maximum suppression, specifically eliminates overlapping bounding boxes and retains only the bounding box with the highest score, and outputs the target bounding box and category information.

5. A non-contact rotating body vibration detection method according to claim 4, characterized in that: Step D: denoise the target bounding box and category information using the SVD denoising method to obtain vibration displacement information. The specific steps are as follows: Step D1: extract the displacement signals in the x-direction and y-direction from the image containing the target bounding box and category information and construct them into binary recursive matrices A respectively. i-1 , as shown in formula (5), Step D2, perform SVD processing on formula (5), and the processing result is shown in formula (6), A i =U i D i V i T (6) Among them, U i (u i1 ,u i2 ) is the left singular matrix obtained by the i-th decomposition, V i (v i1 ,v i2 ) is the right singular matrix obtained by the i-th decomposition, D i is a diagonal matrix; Step D3, after the i-th SVD processing, we get the formula (7): A i =s ai you i1 v i1 T +s bi you i2 v i2 T (7) Among them, v i1 and v i2 V i sub-element of , σ ai u i1 v i1 T and σ bi u i2 v i2 T are the first-order singular values ​​and the second-order singular values, respectively; Step D4, for the matrix σ ai u i1 v i1 T All elements in the two column vectors are summed and averaged, as shown in formula (8), 6. A non-contact rotating body vibration detection system, wherein the specific detection process of the rotating body vibration detection system is based on the rotating body vibration detection method according to any one of claims 1 to 5, characterized in that: It includes an error correction module, an image acquisition module, a feature extraction module and a noise processing module. The error correction module is used to obtain the internal parameter matrix and the external parameter matrix of the high-speed camera by using Zhang's calibration method and complete the conversion between the world coordinate system and the image coordinate system, thereby completing the preliminary error correction of the high-speed camera; The image acquisition module is used to add feature points to the rotating body and obtain the rotating body after the feature points are added, and then use the corrected high-speed camera to shoot the rotating body after the feature points are added to obtain the rotating body image with the feature points; The feature extraction module is used to extract the image feature points of the obtained rotating body image with feature points using the YOLOv7 algorithm and obtain the target bounding box and category information; The noise processing module is used to denoise the target boundary box and category information using the SVD denoising method and obtain vibration displacement information, thereby completing the non-contact detection operation of the rotating body.

Images