Turbine blade film hole detection method based on AI image recognition

Abstract

The invention provides a turbine blade film hole detection method based on AI image recognition. The turbine blade film hole detection method comprises the following steps: S1, establishing a detection system to obtain a turbine blade image containing a film hole; s2, carrying out AI model iteration training development of gas film hole contour extraction; s3, establishing a theoretical blade film hole model; and S4, measuring the turbine blade. The turbine blade film hole detection method based on AI image recognition has the following advantages that firstly, the measurement speed is high, and batch detection can be achieved; 2, the measurement accuracy is high; 3, the method does not depend on the skill level of an inspector; 4, the enterprise detection cost is reduced; and fifthly, the product quality of the turbine blade is powerfully guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of turbine blade film hole detection, and in particular to a turbine blade film hole detection method based on AI image recognition. Background Art

[0002] In the field of turbine blade film hole detection, film holes refer to circular or irregular holes on the surface of high-pressure turbine blades, with a diameter range of 0.3 to 1.5 mm and a depth of <3 mm. The cooling gas extracted from the compressor flows out through the film holes of the turbine blades and adheres to the outer surface of the blades to form an air film, achieving high-temperature isolation and preventing the blades from being ablated by high-temperature combustion gas. AI image recognition refers to the use of computers to extract, analyze and understand target information from extracted images to identify targets and objects of various different patterns.

[0003] Figure 1 Schematic diagram of the air film hole of the turbine blade. Figure 1 As shown, there are hundreds of air film holes distributed on the surface of the turbine blade. The air film holes have small apertures, are irregular holes, and are distributed on the curved surface. It is difficult to evaluate their position and size using traditional three-coordinate measurement or customized measurement tooling.

[0004] At present, in the inspection and acceptance of turbine blades at the manufacturing end, the air film hole position detection is carried out by comparing it with the standard samples through visual inspection to determine whether it is qualified.

[0005] Existing image vision recognition technology uses computers to extract, analyze and understand target information from extracted images in order to identify targets and objects of various different patterns. It has been widely used in application fields such as part surface defect recognition, face recognition, and autonomous driving.

[0006] Currently, visual inspection is used to detect the position of air film holes, which has many disadvantages. For example, there are a large number of air film holes, which are easy to miss, the inspector's subjectivity is strong, and it depends on the inspector's experience. The inspection results are easily affected by the surrounding light intensity and the placement of parts, which can easily lead to misjudgment of the inspection results and cause the quality of parts to escape, posing a potential risk to the performance and stable operation of the engine.

[0007] In view of this, the inventors of the present application designed a turbine blade film hole detection method based on AI image recognition, in order to overcome the above-mentioned technical problems. Summary of the invention

[0008] The technical problem to be solved by the present invention is to overcome the defects of the prior art method for detecting the position of air film holes, such as a large number of air film holes, easy missed detection, many affected factors, and easy misjudgment of results, and to provide a turbine blade air film hole detection method based on AI image recognition.

[0009] The present invention solves the above technical problems through the following technical solutions:

[0010] A method for detecting film holes on turbine blades based on AI image recognition, wherein the method comprises the following steps:

[0011] S1. Build a detection system to obtain images of turbine blades containing air film holes;

[0012] S2. Iterative training and development of AI model for air film hole contour extraction;

[0013] S3, establishing a theoretical leaf film hole model;

[0014] S4. Measure the turbine blades.

[0015] According to one embodiment of the present invention, the detection system in step S1 includes a detection workbench, an adjustable fixture, a light source, a visual device and a turbine blade detection system; the adjustable fixture and the visual device are installed on the detection workbench, and the turbine blade detection system is connected to the visual device.

[0016] According to one embodiment of the present invention, the turbine blade detection system is provided with image recognition and analysis software for identifying image information collected by a visual device.

[0017] According to one embodiment of the present invention, step S2 includes:

[0018] S 21 , positioning the sample high-pressure turbine blade on the test bench by means of an adjustable fixture;

[0019] S 22 2. Obtain the leaf film hole image through visual equipment, and annotate the film hole feature pixels through image segmentation and film hole feature semantic segmentation to establish an AI training data set;

[0020] S 23 , train the model, and after multiple iterations, the detection system can finally determine the characteristics of the air film holes in the picture and automatically frame the outline of the air film holes in the recognition results.

[0021] According to one embodiment of the present invention, the visual device is an image acquisition lens or a camera.

[0022] According to one embodiment of the present invention, the step S 23 The process includes: identifying film holes and extracting image feature information, establishing computer film hole data, and then using AI machine learning to continuously iterate the training model to form a standard turbine blade film hole acceptance algorithm model.

[0023] According to one embodiment of the present invention, step S3 includes: establishing a digital model according to the constructed detection system, importing the theoretical model of the blade to be detected, and obtaining the mapped theoretical two-dimensional image of the air film holes and the air film hole position information of the theoretical model under different fields of view.

[0024] According to one embodiment of the present invention, step S4 includes:

[0025] S 41 , fix the turbine blade to be inspected on the inspection table and use the system to inspect it;

[0026] S 42 , select the information from the AI ​​system and compare it with the air film hole information in the two-dimensional theoretical image of the corresponding field of view;

[0027] S 43 , the comparison results are output. When the position exceeds the preset boundary of the system, it is displayed in red, indicating that the measurement result exceeds the limit; otherwise, it is green, which means it is qualified, assisting or replacing the inspection personnel to complete the qualified judgment of the air film hole position.

[0028] According to one embodiment of the present invention, the step S 42 The system obtains the air film hole contour position and compares it with the standard air film hole theoretical model.

[0029] The positive and progressive effects of the present invention are:

[0030] The turbine blade film hole detection method based on AI image recognition of the present invention has the following advantages:

[0031] 1. Fast measurement speed, batch testing can be realized;

[0032] 2. High measurement accuracy;

[0033] 3. It is not dependent on the inspector’s skill level;

[0034] 4. Reduce enterprise testing costs;

[0035] 5. Effectively guarantee the product quality of turbine blades. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The above and other features, properties and advantages of the present invention will become more apparent through the following description in conjunction with the accompanying drawings and embodiments, in which the same reference numerals always represent the same features, wherein:

[0037] Figure 1 Schematic diagram of the air film hole of the turbine blade.

[0038] Figure 2It is a schematic diagram of the principle of turbine blade film hole detection in the turbine blade film hole detection method based on AI image recognition of the present invention.

[0039] Figure 3 It is a structural schematic diagram of a turbine blade film hole detection system in a turbine blade film hole detection method based on AI image recognition of the present invention.

[0040] Figure 4 It is a schematic diagram of the turbine blade film hole detection structure in the turbine blade film hole detection method based on AI image recognition of the present invention.

[0041] [Reference Signs]

[0042] Testing bench 10

[0043] Adjustable clamp 20

[0044] Light source 30

[0045] Visual equipment 40

[0046] Turbine Blade Inspection System 50

[0047] High pressure turbine blades 60 DETAILED DESCRIPTION

[0048] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0049] Embodiments of the present invention will now be described in detail with reference to the accompanying drawings. Reference will now be made in detail to preferred embodiments of the present invention, examples of which are shown in the accompanying drawings. Wherever possible, the same reference numerals will be used throughout the drawings to represent the same or similar parts.

[0050] Furthermore, although the terms used in the present invention are selected from well-known and commonly used terms, some terms mentioned in the present invention specification may be selected by the applicant at his or her discretion, and their detailed meanings are explained in the relevant parts of the description of this document.

[0051] Furthermore, it is required that the present invention be understood not only by the actual terms used but also by the meanings connoted by each term.

[0052] like Figures 2 to 4 As shown, the present invention discloses a method for detecting air film holes of turbine blades based on AI image recognition, which comprises the following steps:

[0053] Step S1, constructing a detection system to obtain an image of a turbine blade including air film holes.

[0054] Preferably, the detection system in step S1 comprises a detection workbench 10, an adjustable fixture 20, a light source 30, a visual device 40 and a turbine blade detection system 50. The adjustable fixture 20 and the visual device 40 are installed on the detection workbench 10, and the turbine blade detection system 50 is connected to the visual device 40.

[0055] Image recognition and analysis software is provided in the turbine blade detection system 50 for recognizing image information collected by the visual device.

[0056] Step S2: iteratively train and develop an AI model for extracting air film hole contours.

[0057] Preferably, the step S2 includes:

[0058] Step S 21 , position the sample high-pressure turbine blade 60 on the test bench by means of an adjustable fixture.

[0059] For example, by ensuring the stable and consistent positioning of the blade clamping by the 6-point positioning fixture; correcting the distortion of the visual equipment through the grid distortion test target; and calibrating the resolution of the visual equipment through the resolution test target to obtain the scale of the image pixels and the actual size of the part.

[0060] Step S 22 , obtain the image of the film holes on the leaves through visual equipment, and annotate the feature pixels of the film holes through image segmentation and semantic segmentation of the film holes to establish an AI training data set.

[0061] Step S 23 , train the model, and after multiple iterations, the detection system can finally determine the characteristics of the air film holes in the picture and automatically frame the outline of the air film holes in the recognition results.

[0062] For example, the model is trained through tools such as Baidu PaddlePaddle's AI development platform or Google's Tensorflow development platform, image preprocessing module, and customized U-Net image semantic segmentation AI model. The model accuracy evaluation indicator is IOU.

[0063] The trained film hole feature extraction AI model is used to infer the newly acquired images to obtain the feature pixels of the film holes to be measured, and the contour and centroid of each identified film hole feature pixel area are regarded as the actual contour and position of the film hole to be measured. The theoretical contour and centroid position of each film hole to be measured are established based on the visual system camera model, the digital model of the turbine blade, and the tooling calibration information.

[0064] Another method to establish the theoretical contour and position of each measured air film hole is to use a standard blade, place it on the same tooling to obtain its image, and extract the contour and center of mass of the air film hole. The deviation pixels of the measured air film hole contour and center of mass from the reference (theoretical) contour and center of mass are calculated according to the calibrated image scale to obtain the contour deviation and position distance deviation of the measured air film hole, and complete the measurement of the blade air film hole contour and position.

[0065] The visual device 40 here may preferably be an image acquisition lens or a camera.

[0066] In addition, further preferably, in step S 23 The process includes: identifying film holes and extracting image feature information, establishing computer film hole data, and then using AI machine learning to continuously iterate the training model to form a standard turbine blade film hole acceptance algorithm model.

[0067] For example, preferably, in order to improve the efficiency of AI model training, the collected images are first divided into image blocks of specified sizes. Then a supervised image semantic segmentation dataset is created, which consists of images in pairs of grayscale images and labels, where the label is an image that marks non-air film hole feature pixels in the image with 0 and air film hole feature pixels with 1.

[0068] Then divide the dataset into training dataset and validation dataset. Build a custom AI model based on the U-Net model and combined with the Resnet structure, define a dice loss function to avoid imbalance between positive and negative samples, and use the IoU indicator to evaluate the model accuracy.

[0069] Finally, the AI ​​model built using the dataset training is used on the AI ​​development platform. The trained AI model can be used to infer and identify each air film hole pixel on the newly acquired image. Since the camera model has been deformed and calibrated, and the position between the tooling and the camera has also been calibrated, each image pixel can be converted into a coordinate point for distance measurement. Finally, the contour and position of each air film hole can be calculated using the air film hole feature pixels identified by AI.

[0070] Step S3: establishing a theoretical blade film hole model.

[0071] Preferably, step S3 includes: establishing a digital model according to the constructed detection system, importing the theoretical model of the blade to be detected, and obtaining the mapped theoretical two-dimensional image of the air film holes and the air film hole position information of the theoretical model under different fields of view.

[0072] Step S4: measuring the turbine blades.

[0073] Preferably, step S4 includes:

[0074] Step S 41 , fix the turbine blade to be inspected on the inspection table and use the system to inspect it.

[0075] Step S 42 , select the information from the AI ​​system and compare it with the air film hole information in the two-dimensional theoretical image of the corresponding field of view.

[0076] Here the step S 42 The system also includes: obtaining the air film hole contour position and comparing it with the standard air film hole theoretical model.

[0077] Step S 43 , output the comparison results. When the boundary position exceeds the preset system position, it will be displayed in red, indicating that the measurement result exceeds the limit; otherwise, it will be displayed in green, indicating that the measurement result exceeds the limit, and assisting or replacing the inspection personnel to complete the qualified judgment of the air film hole position. The comparison method is to calculate the intersection and union of each pixel set of the measured air film hole feature area and the corresponding reference air film hole feature area pixel set, and the ratio of the number of intersection pixels to the number of union pixels is used as the measurement parameter index t (the index ranges from 0 to 1). The threshold λ for position judgment is determined according to the scale of camera calibration. If t<λ, it indicates that the deviation of the air film hole position is large. Otherwise, the air film hole position is accepted as qualified, and assisting or replacing the inspection personnel to complete the qualified judgment of the air film hole position.

[0078] According to the above method steps, the turbine blade film hole detection method based on AI image recognition of the present invention provides a new solution for the qualified judgment of film hole position detection based on image vision technology. Figure 3 As shown in the figure, the blade film hole image is obtained through visual equipment, the film holes are identified and the image feature information is extracted, and the computer film hole data is established.

[0079] Then AI machine learning is used to continuously iterate the training model and form a standard turbine blade film hole acceptance algorithm model. When the actual blade to be tested is placed in the detection system, the system obtains the film hole contour position and compares it with the standard film hole theoretical model, and displays the qualified film holes, the film hole images that exceed the theoretical position limit, and the offset (such as Figure 4 shown).

[0080] By simultaneously photographing the air film holes in different blade fields and identifying and judging them, the air film holes in the blades can be automatically or assisted by inspectors to determine whether the positions of the air film holes in the blades are qualified, thereby improving inspection efficiency and reducing the risk of "misjudgment and missed inspection".

[0081] The turbine blade film hole detection method based on AI image recognition can automatically identify the film holes in the blade field of view and mark their outlines. The system automatically compares the air film holes with the stored blade theoretical model, determines the film holes with qualified position or position deviation, and displays them on the image (which can be distinguished by color), to assist manual or fully automatic completion of qualified judgment of the film hole position.

[0082] In summary, the turbine blade film hole detection method based on AI image recognition of the present invention has the following advantages:

[0083] 1. Fast measurement speed, batch testing can be realized;

[0084] 2. High measurement accuracy;

[0085] 3. It is not dependent on the inspector’s skill level;

[0086] 4. Reduce enterprise testing costs;

[0087] 5. Effectively guarantee the product quality of turbine blades.

[0088] For those skilled in the art, the above invention disclosure is only used as an example and does not constitute a limitation of the present application. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements and amendments to the present application. Such modifications, improvements and amendments are suggested in the present application, so such modifications, improvements and amendments still belong to the spirit and scope of the exemplary implementation of the present application.

[0089] At the same time, the present application uses specific words to describe the embodiments of the present application. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, some features, structures or characteristics in one or more embodiments of the present application can be appropriately combined.

[0090] Although the specific embodiments of the present invention are described above, it should be understood by those skilled in the art that these are only examples, and the protection scope of the present invention is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but these changes and modifications all fall within the protection scope of the present invention.

Claims

1. A method for detecting film holes on turbine blades based on AI image recognition, characterized in that: The turbine blade film hole detection method comprises the following steps: S1. Build a detection system to obtain images of turbine blades containing air film holes; S2. Iterative training and development of AI model for air film hole contour extraction; S3, establishing a theoretical leaf film hole model; S4. Measure the turbine blades.

2. The turbine blade film hole detection method based on AI image recognition according to claim 1, characterized in that: The detection system in step S1 includes a detection workbench, an adjustable fixture, a light source, a visual device and a turbine blade detection system; the adjustable fixture and the visual device are installed on the detection workbench, and the turbine blade detection system is connected to the visual device.

3. The turbine blade film hole detection method based on AI image recognition according to claim 2, characterized in that: The turbine blade detection system is provided with image recognition and analysis software for recognizing image information collected by visual equipment.

4. The turbine blade film hole detection method based on AI image recognition according to claim 2, characterized in that: The step S2 includes: S 21 , positioning the sample high-pressure turbine blade on the test bench by means of an adjustable fixture; S 22 2. Obtain the leaf film hole image through visual equipment, and annotate the film hole feature pixels through image segmentation and film hole feature semantic segmentation to establish an AI training data set; S 23 , train the model, and after multiple iterations, the detection system can finally determine the characteristics of the air film holes in the picture and automatically frame the outline of the air film holes in the recognition results.

5. The turbine blade film hole detection method based on AI image recognition according to claim 4, characterized in that: The visual device is an image acquisition lens or a camera.

6. The method for detecting air film holes on turbine blades based on AI image recognition according to claim 2, characterized in that: The step S 23 The process includes: identifying film holes and extracting image feature information, establishing computer film hole data, and then using AI machine learning to continuously iterate the training model to form a standard turbine blade film hole acceptance algorithm model.

7. The method for detecting air film holes on turbine blades based on AI image recognition according to claim 2, characterized in that: The step S3 includes: establishing a digital model according to the constructed detection system, importing the theoretical model of the blade to be detected, and obtaining the theoretical two-dimensional image and the position information of the air film holes of the theoretical model under different fields of view.

8. The method for detecting air film holes on turbine blades based on AI image recognition according to claim 2, characterized in that: The step S4 includes: S 41 , fix the turbine blade to be inspected on the inspection table and use the system to inspect it; S 42 , compare the information selected by the AI ​​system with the air film hole information in the two-dimensional theoretical image of the corresponding field of view; S 43 , the comparison results are output. When the position exceeds the preset boundary of the system, it is displayed in red, indicating that the measurement result exceeds the limit; otherwise, it is green, which means it is qualified, assisting or replacing the inspection personnel to complete the qualified judgment of the air film hole position.

9. The method for detecting air film holes on turbine blades based on AI image recognition according to claim 8, characterized in that: The step S 42 The system obtains the air film hole contour position and compares it with the standard air film hole theoretical model.

Images