Blade deformation detection method based on laser correlation sensor and feature point extraction

Abstract

The invention discloses a blade deformation detection method based on laser correlation sensors and feature point extraction. The method comprises the following steps: mounting an elliptic cylinder baffle plate vertical to the cross section of a hub on the hub of a ducted fan; the transmitting end and the receiving end of the laser correlation sensor are installed on the same straight line, and it is ensured that the straight line is tangent to a hub of the motor fan. A plurality of feature reference points with high-contrast colors are uniformly marked on the leaves; when the laser correlation sensor is triggered, the high-speed camera connected with the laser correlation sensor is used for shooting a static image and a dynamic image of the blade in an external triggering manner; performing similarity screening on the collected images; and calculating a deformation vector by comparing the relative positions of the feature points in the static image and the dynamic image so as to judge the deformation degree of the blade. According to the invention, high-precision and real-time monitoring of the blade deformation of the ducted fan can be realized, the accuracy and efficiency of blade deformation detection are improved, blade deformation abnormity can be found in time, corresponding maintenance measures can be taken, and safe and stable operation of the ducted fan is ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of deformation monitoring of ducted fan blades, and particularly relates to a method for detecting blade deformation based on laser pair sensors and feature point extraction. Background Art

[0002] The blades, which are the core components of a ducted fan and the key parts for driving air flow and generating thrust, directly affect the efficiency and safety of the entire system in terms of their design and performance. During the operation of the fan, the blades not only have to bear the strong wind loads generated by high-speed rotation, but also have to cope with the dynamic stresses generated during startup, braking, and speed change transmitted at the connection with the motor hub. The combined action of these forces causes the blades to vibrate, and the dual effects of vibration and aerodynamic effects cause the blades to deform. If the deformation is not effectively managed, it will have a negative impact on the performance of the ducted fan, and in extreme cases, it may even lead to blade breakage and pose a safety hazard. Therefore, it is crucial to accurately monitor the blade deformation, which helps us understand the deformation conditions of the blades under different rotational speeds, so as to take maintenance measures before potential failures occur and effectively prevent blade fracture accidents.

[0003] However, there are obvious deficiencies in the current monitoring of blade deformation. Traditional monitoring methods often rely on manual inspections and regular detections, which are not only inefficient but also difficult to capture the dynamic deformation information of the blades in the high-speed rotation state. In addition, due to the complexity and concealment of blade deformation, traditional static detection methods often have difficulty accurately assessing the actual deformation degree of the blades, resulting in delayed maintenance measures and the inability to effectively prevent blade fracture accidents.

[0004] Therefore, how to achieve high-precision and real-time monitoring of the deformation of ducted fan blades has become an urgent technical problem to be solved. On the one hand, it is necessary to develop a monitoring technology that can accurately capture the dynamic deformation information of the blades in the high-speed rotation state; on the other hand, this technology also needs to have efficient data processing and analysis capabilities to timely detect abnormal blade deformation and take corresponding maintenance measures to ensure the safe and stable operation of the ducted fan.

[0005] The introduction of high-speed camera technology provides a new idea for solving the above technical problems. With its high frame rate and high-resolution capture capabilities, a high-speed camera can clearly record the dynamic images of the blades in the high-speed rotation state, and through comparison with the images of stationary blades, achieve accurate measurement and real-time monitoring of the blade deformation degree. However, how to effectively utilize high-speed camera technology, combined with advanced image processing algorithms and data analysis methods, to achieve high-precision and real-time monitoring of the deformation of ducted fan blades is still an important direction for current technology research and development. Summary of the Invention

[0006] Aiming at the problem of deformation detection of ducted fan blades, a blade deformation detection method based on laser pair sensors and feature point extraction is proposed. This method combines the triggering of laser pair sensors, the evaluation of structural similarity, and the calculation of feature point vectors to accurately judge the degree of blade deformation.

[0007] The technical solution of the present invention is as follows:

[0008] A blade deformation detection method based on laser pair sensors and feature point extraction, characterized by comprising the following steps:

[0009] Install a baffle on the hub of the ducted fan. The baffle is perpendicular to the hub cross-section and can block the laser beam.

[0010] Install the emitting end and receiving end of the laser pair sensor on the same straight line, ensuring that this straight line is tangent to the hub of the motor fan, so that when the blade rotates to a specific position, the laser pair sensor can be triggered.

[0011] Uniformly mark multiple feature reference points on the blade. The feature reference points are marked with high-contrast colors and are distributed at the leading edge, trailing edge, and middle position of the blade.

[0012] Use an external trigger form for image acquisition through a high-speed camera. The high-speed camera is connected to the laser pair sensor. When the baffle triggers the laser pair sensor, the high-speed camera takes static and dynamic images of the blade.

[0013] Perform similarity screening on the collected images, and select the image with the highest structural similarity as the analysis object.

[0014] Use the SIFT algorithm to extract feature points in the image, and screen out the feature points that match the feature reference points marked on the blade.

[0015] By comparing the relative positions of the feature points in the static image and the dynamic image, calculate the deformation vector of the blade in different states to judge the degree of blade deformation.

[0016] Preferably, the design of the baffle needs to consider lightness to avoid excessive impact on the dynamic balance of the fan, and the fixing method of the baffle needs to be firm enough to prevent it from falling off or being damaged when the fan rotates.

[0017] Preferably, the baffle is an elliptical cylinder. The whole cylinder is perpendicular to the hub cross-section. The long axis direction of the elliptical cylinder is designed to be parallel to the linear velocity direction of the hub rotation, and the short axis of the elliptical cylinder is collinear with the hub diameter.

[0018] Preferably, the trigger signal of the laser beam sensor is used to control the shooting of the high-speed camera. In the case of high-speed rotation, the influence of the sensor trigger delay needs to be considered, and the compensation time is calculated through the rotation speed to ensure the accuracy of shooting.

[0019] Preferably, the image similarity screening includes brightness distortion calculation, contrast distortion calculation, and structural correlation calculation, and the comprehensive structural similarity is obtained through weighted summation to screen out the most similar images.

[0020] Preferably, the extraction of the characteristic reference points includes binarizing the image, using the SIFT algorithm to extract the feature points, and screening out the feature points that match the characteristic reference points marked on the blade.

[0021] Preferably, the degree of blade deformation is judged by calculating the vectors of the feature points in the static image and the dynamic image. The longer the modulus length, the greater the degree of deformation, and the change in direction reflects the twisting or bending of the blade.

[0022] Preferably, the installation position of the high-speed camera needs to ensure that when the baffle triggers the laser beam sensor, the characteristic reference points on the blade can clearly appear in the camera lens.

[0023] Preferably, during the image acquisition process, after the first trigger, it is necessary to wait for the time when the blade rotates to the same position for the second time and then trigger the high-speed camera for acquisition to ensure that clear images of the blade during high-speed rotation are captured.

[0024] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0025] 1) By combining the high-precision trigger of the laser beam sensor and the high-frame-rate capture ability of the high-speed camera, the present invention realizes the accurate capture of the dynamic deformation information of the ducted fan blade in the high-speed rotation state. This real-time monitoring method greatly improves the accuracy and timeliness of deformation detection.

[0026] 2) The characteristic reference points evenly distributed on the blade, especially these points located at the leading and trailing edges and the central position of the blade, can comprehensively and accurately reflect the overall deformation of the blade in different states. This layout not only considers the main stress areas of the blade but also ensures the comprehensiveness and accuracy of deformation detection.

[0027] 3) The SIFT algorithm is used to extract the feature points in the image and screen out the feature points that match the characteristic reference points marked on the blade. This process has a high degree of automation, greatly reducing manual intervention and errors. At the same time, by comparing the relative positions of the feature points in the static image and the dynamic image, the deformation vectors of the blade in different states are calculated, thereby realizing the quantitative evaluation of the degree of blade deformation.

[0028] 4) The baffle is designed as an elliptical cylinder and is perpendicular to the hub cross-section. Its major axis direction is parallel to the linear velocity direction of the hub rotation, and the minor axis is collinear with the hub diameter. This design not only ensures the effective shielding of the laser beam but also reduces the impact on the fan dynamic balance, improving the stability and reliability of the system.

[0029] 5) By real-time monitoring the deformation of the blades, the present invention helps to promptly detect potential abnormal deformations, thereby taking corresponding maintenance measures to effectively prevent the occurrence of blade fracture accidents and improve the safety and stability of the ducted fan. Description of the Drawings

[0030] Figure 1 It is a schematic structural diagram of the baffle, where (a) is the front view and (b) is the back view.

[0031] Figure 2 It is a schematic diagram of calibrating multiple uniform points on the blade with a color of relatively high contrast as the blade feature reference points;

[0032] Figure 3 It is a schematic diagram of the installation position of the opposed laser sensor and the connection line of the camera;

[0033] Figure 4 It is the screening of the structural similarity between the static picture and the dynamic picture.

[0034] Figure 5 It is the screening of feature points.

[0035] Figure 6 It is the screening of the corresponding relationship of feature points. Detailed Embodiment

[0036] The technical solution of the present invention will be described in detail below in conjunction with the drawings and embodiments, but the protection scope of the present invention should not be limited thereby.

[0037] This method is mainly divided into two major parts: hardware installation and software processing.

[0038] I. Hardware Installation:

[0039] 1. Baffle Installation:

[0040] Install a baffle perpendicular to the hub cross-section on the hub. The size of the baffle needs to ensure that it can block the laser beam.

[0041] The design of the baffle needs to consider lightness to avoid excessive impact on the fan dynamic balance. At the same time, the fixing method of the baffle needs to be firm enough to prevent it from falling off or being damaged during the rotation of the fan. The baffle structure is as Figure 1As shown, the baffle is an elliptical cylinder. The entire cylinder is perpendicular to the hub cross-section. The major axis direction of the elliptical cylinder is designed to be parallel to the linear velocity direction of the hub rotation, and the minor axis of the elliptical cylinder is collinear with the hub diameter.

[0042] 2. Installation of the laser pair sensor:

[0043] Install the transmitter and receiver of the laser pair sensor on the same straight line, and ensure that this straight line is tangent to the hub of the motor fan to ensure that when the blade rotates to a specific position, the laser pair sensor can be accurately triggered.

[0044] 3. Blade feature reference points:

[0045] Select the blade that needs to be deformed and detected, and evenly dot small points with a diameter of about 3 mm on the blade. These points are marked with a color with a higher contrast to be used as feature points in subsequent image analysis, as Figure 2 shown.

[0046] These feature points altogether form three lines, which are respectively distributed at the leading edge, trailing edge and middle position of the blade to comprehensively reflect the deformation of the blade. The installation position of the pair laser sensor and the camera connection line are as Figure 3 shown.

[0047] 4. Image acquisition is carried out in an external trigger mode by a high-speed camera, specifically including:

[0048] 4.1 Configuration of the high-speed camera and the laser pair sensor:

[0049] The high-speed camera is installed at a specific position. The selection of this position is to ensure that when the baffle triggers the laser pair sensor, the feature reference points on the blade can clearly appear in the camera lens. The trigger interface of the high-speed camera is connected to the laser pair sensor so that the trigger signal sent by the laser pair sensor can directly control the shooting of the high-speed camera.

[0050] 4.2 Shooting static reference images:

[0051] When the motor is not started, first turn on the laser pair sensor. Then manually turn the fan blade until the baffle triggers the laser pair sensor. At this time, the high-speed camera will shoot a static image of the blade. This image will be used as a reference for subsequent dynamic shooting.

[0052] After shooting the static image, turn off the laser pair sensor, and then start the ducted fan to the specified speed.

[0053] 4.3 Dynamic image acquisition:

[0054] When the fan reaches a stable speed, turn on the laser pair sensor again and adjust the high-speed camera to the acquisition mode.

[0055] Due to the triggering delay of the laser pair sensors, especially in the case of high-speed rotation, this delay phenomenon will be more obvious. Therefore, after the first trigger, it is necessary to wait for the time t when the blade rotates to the same position for the second time and then trigger the high-speed camera for acquisition.

[0056] After the rotational speed stabilizes, the time required for the blade to rotate to the same position is the same and can be calculated based on the rotational speed. Let the rotational speed be k (min / s) and the sensor delay be τ (s), then the camera is triggered for shooting after the time t.

[0057] Therefore, in the process of image acquisition by the high-speed camera using the external trigger mode, it is necessary to accurately configure the parameters of the camera and the sensor, as well as reasonably design the shooting process to ensure that clear images of the blade during high-speed rotation can be captured. At the same time, the influence of the sensor triggering delay needs to be considered and corresponding compensation should be carried out to ensure the accuracy of the shooting results.

[0058] II. Software processing:

[0059] 1. Screening of images with similar angles

[0060] During the process of image acquisition by the high-speed camera, due to the delay of the laser pair sensors, multiple images with similar but slightly different angles will be captured. In order to screen out the most suitable images from these pictures, similarity comparison and screening are required. Specifically, it includes:

[0061] 1) Calculation of image brightness distortion:

[0062] The average gray value of the image reflects the brightness of the image. By calculating the average gray values of two pictures and establishing a brightness comparison function l(a, b), the difference in brightness between the two pictures is quantified.

[0063]

[0064] Among them, C1 is a very small positive real number to prevent the denominator from being 0; μ A is the average gray value of image A, μ B is the average gray value of image B, where a i is the gray value of the i-th pixel in image A, and b i is the gray value of the i-th pixel in image B.

[0065] The value range of the brightness comparison function is usually in (0, 1], and the closer it is to 1, the more similar the two pictures are in terms of brightness.

[0066] 2) Calculation of contrast distortion:

[0067] The contrast of an image is expressed by calculating the standard deviation of the pixel grayscale. The larger the standard deviation, the greater the difference in pixel values ​​in the image, that is, the stronger the contrast.

[0068] A contrast comparison function c(a,b) is established to quantify the difference in contrast between two images.

[0069]

[0070] Where C2 is a very small positive real number that prevents the denominator from being zero; σ A is the grayscale standard deviation of image A,

[0071] σ B is the grayscale standard deviation of image B, The contrast function usually ranges from (0,1].

[0072] 3) Structural correlation calculation:

[0073] Image structural similarity is compared by calculating the correlation of corresponding pixels.

[0074] Similarity coefficient between images A and B:

[0075] Structural Correlation Function Where C3 is a very small positive real number that prevents the denominator from being zero.

[0076] 4) Comprehensive calculation of structural similarity:

[0077] In order to obtain comprehensive structural similarity, different weights can be assigned to brightness, contrast, and structural relevance, and their weighted summation can be used to obtain a value that reflects the overall structural similarity between the two images.

[0078]

[0079] Among them, α, β, and γ are the proportions of different features in the structural similarity.

[0080] 5) Filter the most similar pictures:

[0081] From all the collected images, the one with the highest structural similarity is selected as the final analysis object. This usually involves calculating the similarity of each image with all other images and finding the one with the highest similarity.

[0082]

[0083] Among them, the closer the structural similarity is to 1, the more similar the pictures are. The closer the structural similarity is to 0, the lower the picture similarity. Select the picture with the highest structural similarity. The flow chart is as Figure 4 shown.

[0084] 2. Feature reference point extraction

[0085] 1) Binarization processing:

[0086] To extract feature points more easily, binarize the static picture and the dynamic picture, that is, convert the picture into an image with only two colors, black and white.

[0087] 2) Extract feature points using the SIFT algorithm:

[0088] Use the SIFT (Scale-Invariant Feature Transform) algorithm to extract feature points in the picture. The SIFT algorithm is a computer vision algorithm used to detect and describe local features in images. These feature points have good robustness to scale changes, rotations, illumination changes, etc.

[0089] 3) Screen feature reference points:

[0090] From the extracted feature points, select the part that coincides with the feature reference points drawn on the blade as the analysis object. This usually involves a process of matching and screening feature points, as Figure 5 shown.

[0091] 4) Compare the relative positions of feature points:

[0092] Further compare the feature points selected from the static picture and the dynamic picture, and determine whether the relative positions of the feature points in the two pictures are consistent by the relative distance between points. As Figure 6 shown. If they are consistent, it means that the two pictures have high similarity in feature points.

[0093] 3. Judgment of blade deformation degree

[0094] 1) Calculate vectors:

[0095] Use the feature points in the static picture as the starting points and the feature points in the dynamic picture as the ending points to obtain a set of vectors. These vectors reflect the deformation of the blade in different states.

[0096] 2) Judge the deformation degree:

[0097] Judge the deformation degree of the blade by calculating the magnitude and direction of the vector values in different regions. Among them, the longer the modulus length, the greater the deformation degree; the change in direction can reflect the twisting or bending of the blade.

[0098] 3) Comprehensive analysis:

[0099] Combining the vector information of all feature points, the deformation degree of the blade can be comprehensively analyzed.

[0100] Example:

[0101] (1) Import static pictures and dynamic pictures into Matlab respectively;

[0102] (2) Grayscale all the imported pictures;

[0103] (3) Through the structural similarity algorithm, taking the static picture as the reference picture, compare the dynamic pictures for structural similarity. The comparison returns a value between 0 and 1, and the closer the value is to 1, the higher the similarity. Select the dynamic picture with the highest structural similarity among all dynamic pictures.

[0104] (4) Binarize the static picture and the selected dynamic picture;

[0105] (5) Use the SIFT algorithm to find the 30 feature points with the strongest features in the static picture and the dynamic picture respectively;

[0106] (6) Set a range according to the position of the blade in the picture, and roughly select the feature points according to this range;

[0107] (7) For the remaining feature points, judge whether the pixel points on the ring with a radius of 10 pixels centered on each feature point are black. If they are black, they are judged as feature reference points.

[0108] (8) Set a threshold distance d max = 2, calculate the distance between a feature point in the static picture and all points in the dynamic picture. If the distance is less than the threshold, it is the corresponding point.

[0109] (9) Calculate the vector with the static feature point as the starting point and the corresponding dynamic feature point as the ending point. Obtain the deformation condition of the blade according to the modulus and direction of the vector, where the longer the modulus, the greater the deformation degree.

[0110] A method for detecting blade deformation based on laser pair sensors and feature point extraction proposed by the present invention has technical effects such as high precision, real-time monitoring, comprehensively reflecting blade deformation, efficient data processing and analysis, optimizing sensor design, and strong adaptability.

Claims

1. A blade deformation detection method based on laser beam sensor and feature point extraction, characterized in that: The following steps are involved: A baffle is installed on the hub of the ducted fan, wherein the baffle is perpendicular to the cross section of the hub and can block the laser beam; Install the transmitting end and receiving end of the laser beam sensor on the same straight line, and ensure that the straight line is tangent to the hub of the motor fan, so that the laser beam sensor can be triggered when the blade rotates to a specific position; Uniformly marking a plurality of characteristic reference points on the blade, wherein the characteristic reference points are marked with high-contrast colors and are distributed at the leading edge, the trailing edge and the middle position of the blade; The image is collected by a high-speed camera in an external triggering form. The high-speed camera is connected to a laser beam sensor. When the baffle triggers the laser beam sensor, the high-speed camera takes a static image and a dynamic image of the blade. Perform similarity screening on the collected images and select the images with the highest structural similarity as the analysis objects; The SIFT algorithm is used to extract feature points in the image, and the feature points that match the feature reference points marked on the leaves are selected; By comparing the relative positions of feature points in static images and dynamic images, the deformation vector of the blade in different states is calculated to determine the deformation degree of the blade.

2. The blade deformation detection method based on laser beam sensor and feature point extraction according to claim 1 is characterized in that: The baffle is an elliptical cylinder, the entire cylinder is perpendicular to the hub cross section, the major axis direction of the elliptical cylinder is designed to be parallel to the linear velocity direction of the hub rotation, and the minor axis of the elliptical cylinder is colinear with the hub diameter.

3. The blade deformation detection method based on laser beam sensor and feature point extraction according to claim 1 is characterized in that: The trigger signal of the laser beam sensor is used to control the shooting of the high-speed camera. In the case of high-speed rotation, the influence of the sensor trigger delay needs to be considered, and the compensation time is calculated by the rotation speed to ensure the accuracy of the shooting.

4. The blade deformation detection method based on laser beam sensor and feature point extraction according to claim 1 is characterized in that: The image similarity screening includes brightness distortion calculation, contrast distortion calculation and structural correlation calculation, and a comprehensive structural similarity is obtained by weighted summation to screen out the most similar images.

5. The blade deformation detection method based on laser beam sensor and feature point extraction according to claim 1 is characterized in that: The extraction of the characteristic reference points includes binarizing the image, extracting the characteristic points using the SIFT algorithm, and screening out the characteristic points that match the characteristic reference points marked on the leaf.

6. The blade deformation detection method based on laser beam sensor and feature point extraction according to claim 1 is characterized in that: The blade deformation degree is determined by calculating the vectors of the feature points in the static image and the dynamic image. The longer the modulus is, the greater the deformation degree is, and the change in direction reflects the twisting or bending of the blade.

7. The blade deformation detection method based on laser beam sensor and feature point extraction according to claim 1 is characterized in that: The installation position of the high-speed camera must ensure that when the baffle triggers the laser beam sensor, the characteristic reference point on the blade can clearly appear in the camera lens.

8. The blade deformation detection method based on laser beam sensor and feature point extraction according to claim 1 is characterized in that: During the image acquisition process, after the first trigger, it is necessary to wait for the blade to rotate to the same position for the second time before triggering the high-speed camera for acquisition, so as to ensure that a clear image of the blade during high-speed rotation is captured.

9. The blade deformation detection method based on laser beam sensor and feature point extraction according to claim 1 is characterized in that: The high contrast color marking uses fluorescent materials to draw reference points on the blades, and ultraviolet light is used to illuminate these reference points.

Images