A Smart Analysis Method for Hole Exploration Videos of Aero-engine Rotors

Abstract

This invention relates to the field of aero-engine damage detection (IPC classification: G06T), and particularly to an intelligent analysis method for aero-engine rotor borehole inspection videos. The intelligent analysis method for aero-engine rotor borehole inspection videos comprises a data acquisition module, a region selection module, a direction determination module, a hash similarity calculation module, a full rotation determination module, a tiling and merging module, and a coating peeling ratio calculation module. This invention automatically determines the rotor's motion direction in the rotor video and uses normalized hash similarity to determine whether the rotor has rotated one full rotation. This process is all performed automatically by a computer, resulting in a high degree of automation and accuracy. Furthermore, the coating peeling ratio obtained by directly using manual annotation and statistical methods is highly accurate, far exceeding the recognition accuracy of traditional methods relying on visual estimation.

Description

Technical Field

[0001] This invention relates to the field of aero-engine damage detection, IPC classification number: G06T, and particularly to an intelligent analysis method for aero-engine rotor borehole inspection videos. Background Technology

[0002] Borehole inspection is a non-destructive testing technique that allows for the inspection of the internal condition of an aircraft engine without disassembling it. During the inspection, personnel insert a borehole probe into the engine through a pre-drilled hole and transmit an optical image of the engine's interior via a connected fiber optic cable. The obtained image is then used to assess the damage to the engine's internal structure. Because borehole inspection eliminates the need for engine disassembly, it avoids the costs and secondary damage associated with disassembly. Therefore, borehole inspection plays a crucial role in the health monitoring of aircraft engines; statistics show that over 90% of internal aircraft engine faults are detected using this technique.

[0003] High-pressure rotor duct coating is a coating sprayed onto the high-pressure rotor ducts of aero engines. Its main function is to prevent damage to heat-sensitive components inside the rotor from the high-temperature environment of the duct. Therefore, the high-pressure rotor duct coating is of great significance for ensuring the normal operation of aero engines. However, as aero engines age, the high-pressure rotor duct coating will gradually peel off under the erosion of airflow. When the proportion of coating peeling exceeds a certain range, the high-pressure rotor duct coating will lose its protective function for heat-sensitive components inside the aero engine rotor. Therefore, regular borehole inspection is necessary to monitor the peeling of the high-pressure rotor duct coating.

[0004] Current detection methods involve capturing video of the high-pressure rotor duct coating rotating once using a borescope, then visually estimating the coating peeling percentage. This estimation method has a significant margin of error. However, aviation maintenance manuals often specify precise limits for the high-pressure rotor duct coating peeling percentage. For example, for a certain stage of the V2500, the peeling percentage must not exceed 40%; exceeding 40% necessitates factory return for repairs, which can cost 1.5-2 million NTD. Therefore, current methods cannot accurately calculate the high-pressure rotor duct coating peeling percentage, resulting in significant errors and disputes in maintenance decisions. If the actual peeling percentage exceeds the specified threshold, allowing the engine to continue operating if the inspector estimates it to be below the threshold could impact flight safety. Conversely, if the peeling percentage is below the threshold but maintenance personnel estimate it to be above, it would lead to substantial wasted maintenance costs. Therefore, accurately calculating the coating peeling percentage is crucial for ensuring flight safety and reducing maintenance costs.

[0005] Therefore, the industry urgently needs a technology to accurately calculate the peeling rate of high-pressure rotor duct coatings, thereby ensuring correct maintenance decisions, reducing maintenance costs, and guaranteeing the operational safety of aero engines. Furthermore, the technology for accurately calculating the peeling rate of high-pressure rotor duct coatings can be used to study quantitative indicators such as the peeling rate of new technologies after modifications to coating formulations or processes, thus better evaluating these new technologies. Summary of the Invention

[0006] The purpose of this invention is to extract a complete planar image of the rotor air passage coating surface from the rotor borehole inspection video, and to accurately calculate the coating peeling ratio in the planar image, so as to provide a reference for the maintenance decision of aero-engines and solve the safety problems and waste of maintenance costs caused by the large errors of traditional detection methods.

[0007] The technical solution adopted by this invention to solve its technical problem is:

[0008] A smart analysis method for borehole inspection videos of aero-engine rotors comprises a data acquisition module, a region selection module, a direction determination module, a hash similarity calculation module, a full rotation determination module, a tiling and merging module, and a coating peeling ratio calculation module.

[0009] Preferably, the data acquisition module uses an industrial endoscope to acquire borehole video of the rotor inside the aero-engine.

[0010] Preferably, the region selection module crops the borehole video into multiple smaller rectangular regions. Preferably, the hash similarity calculation module uses a difference hash algorithm to determine the similarity between any two images of the planar image.

[0011] Preferably, the direction determination module is used to determine the direction of rotor movement in the borehole video.

[0012] Preferably, the whole-circle rotation determination module is used to determine the hash similarity between the selected area video and the initial frame.

[0013] Preferably, the tiling merging module uses hash similarity to tile and merge all frames from the initial frame to the stop frame.

[0014] Preferably, the coating peeling ratio calculation module is used to calculate the coating peeling ratio.

[0015] More preferably, the data acquisition module specifically includes: placing an industrial endoscope inside the rotor of an aero-engine to acquire rotor borehole video data in real time, and segmenting it into borehole image data frame by frame.

[0016] More preferably, the region selection module selects and crops out multiple smaller rectangular regions from the rotor borehole video, and generates corresponding rotor planar images for the cropped regions.

[0017] More preferably, the hash similarity calculation module specifically includes: processing the generated planar image using a difference hash algorithm to generate the image's hash value, and using normalized hash similarity to determine the similarity between any two images of the planar image; the higher the hash similarity, the higher the similarity between the two images; the calculation formula is as follows:

[0018]

[0019] In the formula S hash Let d(hash1, hash2) be the normalized hash similarity between image 1 and image 2, and let len ​​be the Hamming distance between the hash values ​​of image 1 and image 2. hash1 Let be the hash value length of image 1. When the hash value lengths of image 1 and image 2 are the same, that is:

[0020] len hash1 =len hash2

[0021] More preferably, the direction determination module determines the direction of rotor movement in the rotor video, and rotates the video according to the direction of movement so that the rotor movement direction is from right to left.

[0022] More preferably, the whole-cycle rotation determination module determines whether the video has rotated one full cycle by comparing the hash similarity between each rectangular region after region selection and the initial frame.

[0023] More preferably, the tiling and merging module performs tiling and merging of the initial frame and all frames from the initial frame to the stop frame using hash similarity to obtain a planar image of the entire circumference of the high-pressure rotor air passage spraying surface.

[0024] More preferably, after the rotor planar image obtained by the coating peeling ratio calculation module is tiled and merged, the corresponding coating peeling ratio can be obtained by manually annotating and statistically analyzing the proportion of the pixels in the coating peeling area to the total pixels in the rotor planar image.

[0025] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0026] 1. This invention automatically determines the direction of rotor movement in the rotor video and uses normalized hash similarity to determine whether the rotor has rotated one revolution. This process is all determined and calculated by the computer itself, with a high degree of automation and high accuracy.

[0027] 2. This invention automatically obtains a planar image of the complete circumference of the rotor coating surface by using a tiling and merging method. The obtained planar image of the rotor coating surface is smooth and free from splicing defects such as discontinuities and jumps that are visible to the naked eye, resulting in a good splicing effect.

[0028] 3. The present invention selects a small area during the tiling and merging process, with no obvious deformation. In addition, the number of pixels spliced ​​by the tiling and merging method is small, and there will be no splicing error due to the distortion of the image in the video. The resulting spliced ​​planar image is basically consistent with the actual unfolded planar image.

[0029] 4. The coating peeling ratio obtained by the present invention through the method of directly using manual annotation to statistically determine the coating peeling ratio has a very high accuracy, far exceeding the identification accuracy of the traditional method that relies on visual estimation. It can correctly assess the coating peeling damage of the aero-engine rotor air passage and make correct maintenance decisions. Attached Figure Description

[0030] The embodiments of this utility model will be described more fully with reference to the accompanying drawings. However, the accompanying drawings are for illustration and explanation only and do not constitute a limitation on the scope of this utility model.

[0031] Figure 1 This is a flowchart of an intelligent analysis method for borehole inspection videos of an aero-engine rotor according to the present invention.

[0032] Figure 2 This is a flowchart of the direction determination module of an intelligent analysis method for borehole inspection video of an aero-engine rotor according to the present invention.

[0033] Figure 3 This invention relates to a whole-cycle rotation determination module for an intelligent analysis method of borehole inspection video of an aero-engine rotor.

[0034] Figure 4 This invention relates to a tiling and merging module for an intelligent analysis method of borehole exploration videos of an aero-engine rotor. Detailed Implementation

[0035] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the present invention.

[0036] In this embodiment, an intelligent analysis method for borehole inspection videos of aero-engine rotors includes a data acquisition module, a region selection module, a direction determination module, a hash similarity calculation module, a full rotation determination module, a tiling and merging module, and a coating peeling ratio calculation module. The main steps are S1 to S7, as follows: Figure 1 As shown, the details are as follows:

[0037] Step S1: Use an industrial endoscope to collect real-time borehole inspection video of the rotor inside the aero-engine and segment it into borehole image data frame by frame.

[0038] Step S2: Select and crop out multiple smaller rectangular areas from the rotor borehole video, and generate corresponding rotor planar images for the cropped areas;

[0039] Step S3: Process the generated rotor plane image based on the differential hashing algorithm to generate the image's hash value, and use normalized hash similarity to determine the similarity between any two images of the plane image; the calculation formula is as follows:

[0040]

[0041] In the formula S hash Let d(hash1, hash2) be the normalized hash similarity between image 1 and image 2, and let len ​​be the Hamming distance between the hash values ​​of image 1 and image 2. hash1 Let be the hash value length of image 1. When the hash value lengths of image 1 and image 2 are the same, that is:

[0042] len hash1 =len hash2

[0043] The higher the hash similarity, the higher the similarity between the two images. The hash similarity algorithm is used to determine the rotor orientation and full rotation.

[0044] Step S4: Determine the direction of rotor motion in the rotor video. The determination process is as follows: Figure 2 As shown:

[0045] Step 4-1: Based on steps S1 and S2, set the initial frame position, and determine the hash similarity Shash-1 between the Kth frame and the initial frame based on the hash similarity algorithm; if the hash similarity Shash-1 value is greater than the threshold, then perform a comparison of K+1 frames; if the hash similarity Shash-1 value is less than the threshold, then enter the rotation degree recognition process; until the stop frame ends.

[0046] Step 4-2: Based on Step 4-1, the initial frame image is judged. If the horizontal length is less than the vertical length, the initial frame image is rotated by an angle a = 0; if the horizontal length is greater than the vertical length, the initial frame image is rotated by an angle a = 90°. The image of the Kth frame that entered the recognition process in Step 4-1 is rotated by an angle a. The hash value of each column of pixels in the Kth frame image is compared with the middle hash value of the initial frame image to obtain the position with the minimum hash similarity.

[0047] Step 4-3: Compare the position with the minimum hash similarity obtained in Step 4-2 with the position in the middle column. If it is on the left, rotate the position with the minimum hash similarity by an angle 'a'; if it is on the right, rotate the position with the minimum hash similarity by an angle 'a' = 'a' + 180'.

[0048] The rotor rotates in the direction shown in the video.

[0049] Rotate the video according to the direction of rotor motion, so that the rotor motion direction is from right to left.

[0050] Step S5: Determine the hash similarity between the video in the region and the initial frame. The determination process is as follows: Figure 3 As shown:

[0051] Step S5-1: Based on steps S1 and S2, set the initial frame position, and determine the hash similarity Shash-2 between the Kth frame and the initial frame based on the hash similarity algorithm. If the Shash-2 value is greater than the set threshold, then perform a comparison of frame K+1; if the Shash-2 value is less than the set threshold, then enter the full rotation recognition process; until the stop frame ends.

[0052] Step S5-2: Based on step S5-2, compare the hash similarity between the image of the Kth frame and the image of the initial frame. If it is greater than the set threshold, proceed to the next step to determine whether K is less than the total number of video frames. If so, compare K+1 frames. If it is less than the set threshold, rotate one full circle and stop at frame position S. If K in the previous step is greater than the total number of video frames, rotate one full circle and stop at frame position S.

[0053] Step S6: Based on steps S4 and S5, the last column of pixels in the initial frame is compared with the hash similarity of the Kth frame image column by column. The position P corresponding to the minimum hash similarity value can be determined. After stitching the pixels after column P of the Kth frame to the merged planar image, it is determined whether K is less than S. If it is less, K+1 comparison is performed; if it is greater, a rotor planar image is formed.

[0054] Step S7: After obtaining the rotor plane image in step S6, the corresponding coating peeling ratio can be obtained by manually annotating and statistically analyzing the proportion of pixels in the coating peeling area to the total pixels of the rotor plane image, so as to make the correct maintenance decision.

Claims

1. An intelligent analysis method for borehole inspection videos of aero-engine rotors, characterized in that, It consists of a data acquisition module, a region selection module, a direction determination module, a hash similarity calculation module, a full rotation determination module, a tiling and merging module, and a coating peeling ratio calculation module. The data acquisition module uses an industrial endoscope to collect borescope videos of the rotor inside an aero-engine. The region selection module crops the borehole video into multiple smaller rectangular regions; The hash similarity calculation module uses the difference hash algorithm to determine the degree of similarity between any two images of the planar image; The direction determination module is used to determine the direction of rotor movement in the borehole video; The whole-circle rotation determination module is used to determine the hash similarity between the selected area video and the initial frame; The tile merging module uses hash similarity to tile and merge all frames from the initial frame to the stop frame. The coating peeling ratio calculation module is used to calculate the coating peeling ratio.

2. The intelligent analysis method for borehole inspection video of an aero-engine rotor according to claim 1, characterized in that, The data acquisition module specifically includes: placing an industrial endoscope inside the rotor of an aero-engine to acquire rotor borehole video data in real time, and segmenting it into borehole image data frame by frame.

3. The intelligent analysis method for borehole inspection video of an aero-engine rotor according to claim 1, characterized in that, The region selection module selects and crops out multiple smaller rectangular regions from the rotor borehole video, and generates corresponding rotor planar images for the cropped regions.

4. The intelligent analysis method for borehole inspection video of an aero-engine rotor according to claim 1, characterized in that, The hash similarity calculation module specifically includes: processing the generated planar image using a difference hash algorithm to generate the image's hash value, and using normalized hash similarity to determine the similarity between any two images of the planar image. The calculation formula is as follows: In the formula S hash Let d(hash1, hash2) be the normalized hash similarity between image 1 and image 2, and let len ​​be the Hamming distance between the hash values ​​of image 1 and image 2. hash1 Let be the hash value length of image 1. When the hash value lengths of image 1 and image 2 are the same, that is: len hash1 =len hash2 The higher the hash similarity, the more similar the two images are.

5. The intelligent analysis method for borehole inspection video of an aero-engine rotor according to claim 1, characterized in that, The direction determination module determines the direction of rotor movement in the rotor video.

6. The intelligent analysis method for borehole inspection video of an aero-engine rotor according to claim 5, characterized in that, The rotor rotates in the direction shown in the video.

7. The intelligent analysis method for borehole inspection video of an aero-engine rotor according to claim 1, characterized in that, The full rotation determination module determines whether the video has rotated one full circle by comparing the hash similarity between each rectangular region selected and the initial frame.

8. The intelligent analysis method for borehole inspection video of an aero-engine rotor according to claim 1, characterized in that, The tiling and merging module performs tiling and merging of the initial frame and all frames from the initial frame to the stop frame using hash similarity to obtain a planar image of the entire circumference of the high-pressure rotor air passage spraying surface.

9. The intelligent analysis method for borehole inspection video of an aero-engine rotor according to claim 1, characterized in that, The rotor planar diagram obtained by merging the coating peeling ratio calculation module.

10. The intelligent analysis method for borehole inspection video of an aero-engine rotor according to claim 8, characterized in that, The coating peeling ratio is obtained by manually annotating and statistically analyzing the proportion of pixels in the peeling area to the total number of pixels in the rotor planar image.

Images