CN114782313A - Statistical method for high-temperature alloy additive manufacturing cracks - Google Patents
Statistical method for high-temperature alloy additive manufacturing cracks Download PDFInfo
- Publication number
- CN114782313A CN114782313A CN202210258918.0A CN202210258918A CN114782313A CN 114782313 A CN114782313 A CN 114782313A CN 202210258918 A CN202210258918 A CN 202210258918A CN 114782313 A CN114782313 A CN 114782313A
- Authority
- CN
- China
- Prior art keywords
- additive manufacturing
- picture
- temperature alloy
- cracks
- superalloy
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10056—Microscopic image
- G06T2207/10061—Microscopic image from scanning electron microscope
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30136—Metal
Abstract
The invention belongs to the technical field of high-temperature alloy additive manufacturing, and discloses a statistical method for high-temperature alloy additive manufacturing cracks. The core part of the method is that a Unet model and a convolutional neural network model of a semantic segmentation algorithm are utilized to automatically process a metallographic structure picture obtained from an additive manufacturing high-temperature alloy sample, the average crack rate of the additive manufacturing high-temperature alloy sample is rapidly and accurately obtained, and therefore the establishment of the evaluation method of the crack sensitivity of the additive manufacturing high-temperature alloy is promoted.
Description
Technical Field
The invention belongs to the technical field of high-temperature alloy additive manufacturing, and particularly relates to a statistical method for high-temperature alloy additive manufacturing cracks.
Background
The high-temperature alloy material has excellent high-temperature resistance, fatigue resistance, oxidation resistance and corrosion resistance, is mainly applied to the manufacture of core hot end parts of aero-engines and gas turbines, and is widely applied to the fields of aerospace, energy power and the like. The additive manufacturing technology meets the requirement of manufacturing hollow turbine blades with complex inner cavity structures, but because the content of alloy elements of high-performance high-temperature alloy is high, the defects of solidification cracks, liquefaction cracks and the like are easy to occur in the additive manufacturing process, and great obstruction is brought to the additive manufacturing production of advanced high-temperature alloy turbine blades. Aiming at the problem of cracks in high-temperature alloy additive manufacturing, the cracks are counted and compared, and the design of the high-temperature alloy for low-crack additive manufacturing is facilitated to be accelerated.
At present, the domestic statistical method for the high-temperature alloy additive manufacturing cracks comprises the following steps:
(1) and (3) an artificial binary segmentation method. Selecting crack parts on the metallographic picture by using Image processing tools such as Photoshop and the like, deleting or filling ground color in non-crack regions to distinguish the crack regions from the non-crack regions, then performing binary segmentation by using metallographic picture processing software such as Image Pro Plus and the like, and calculating the ratio of the area of the crack regions to the total area of the picture, namely the crack rate;
(2) automatic binary segmentation method. The method comprises the following steps of performing automatic binary segmentation on a metallographic picture by using an existing algorithm to segment a crack region and a non-crack region, wherein manual intervention is required at present on the accuracy of region segmentation, and air holes generated in the material increase manufacturing process and scratches and the like generated due to improper operation in the grinding and polishing process are easily scratched into the crack region by the automatic binary segmentation;
(3) XCT test method. The sample is tested and analyzed by using a three-dimensional X-ray tomography scanning device, the spatial positions and sizes of cracks and pores can be accurately calculated, and the crack rate of the sample can be calculated according to the ratio of the crack volume to the size of the test sample;
the above three methods, although capable of counting cracks in the additive-manufactured superalloy, exhibit several problems as follows: the artificial binary segmentation method is subjectively influenced, andthe automatic binary segmentation method is easy to count pores, scratches and the like into crack regions, so that the crack rate is large, and the XCT method is accurate in test, but the size of a test sample is limited to 1-2 mm due to the limitation of the penetrating power of an X-ray source3And the crack rate statistics of large-batch samples cannot be met.
Therefore, in view of the problems of the above methods, it is necessary to develop a new statistical method for additive manufacturing cracks of high temperature alloys.
Disclosure of Invention
The invention aims to provide a statistical method for high-temperature alloy additive manufacturing cracks, aiming at the defects of the prior art. The core part of the method is that a Unet model and a convolutional neural network model of a semantic segmentation algorithm are utilized to automatically process a metallographic structure picture obtained from an additive manufacturing high-temperature alloy sample, the average crack rate of the additive manufacturing high-temperature alloy sample is rapidly and accurately obtained, and therefore the establishment of the evaluation method of the crack sensitivity of the additive manufacturing high-temperature alloy is promoted.
In order to achieve the above object, the present invention provides a statistical method for high temperature alloy additive manufacturing cracks, which comprises the following steps:
s1: performing laser additive manufacturing on the high-temperature alloy powder by using an additive manufacturing system to obtain an additive manufactured high-temperature alloy sample; cutting the additive manufacturing high-temperature alloy sample, and grinding and polishing to obtain a plurality of sections; shooting the whole area of each section after polishing by using a metallographic microscope to obtain a corresponding original metallographic picture;
s2: establishing an automatic preprocessing program of the picture file; cutting each original metallographic picture into a plurality of sub-pictures with the same size by using the picture file automatic preprocessing program;
s3: writing a Unet model and a convolutional neural network model of a semantic segmentation algorithm; performing semantic segmentation on the sub-picture obtained in the step S2 by using the Unet model of the semantic segmentation algorithm, determining cracks, scratches and pore parts, and automatically identifying the cracks, the scratches and the pores by using the convolutional neural network model;
s4: and counting the pixel points of the crack parts and the total pixel points of each original metallographic picture, calculating the crack rate of each original metallographic picture by using a formula, and then calculating the crack rate of the high-temperature alloy sample manufactured by the additive manufacturing method.
According to the present invention, preferably, in step S1, the additive manufacturing system is a laser co-axial powder feeding additive manufacturing system or a laser powder bed additive manufacturing system.
According to the invention, in step S1, the superalloy powder is prepared by argon atomization of a superalloy substrate, and the particle size of the superalloy powder is 80-250 meshes.
According to the present invention, preferably, in step S1, the cutting is wire cutting the additive manufacturing superalloy sample with a wire cutting machine.
According to the present invention, in step S1, the magnification of the metallographic microscope is preferably 100 times or more.
According to the present invention, preferably, the original metallographic picture adopts an RGB mode.
According to the present invention, preferably, in step S2, the picture file automatic preprocessing program is created using the os library and PIL library functions based on python.
According to the present invention, preferably, in step S2, the picture file automatic preprocessing program includes picture file loading, picture file cropping, picture renaming, and picture file modular storage.
According to the invention, preferably, the size of the sub-picture is 64 pixels by 64 pixels, 128 pixels by 128 pixels or 256 pixels by 256 pixels.
According to the present invention, preferably, in step S3, the uet model and the convolutional neural network model of the semantic segmentation algorithm are each written independently by a keras library function of python.
According to the present invention, preferably, the convolutional neural network model comprises a convolutional layer, a pooling layer and a full-link layer, and further preferably, the convolutional neural network model comprises four convolutional layers, four pooling layers and a full-link layer.
According to the present invention, preferably, the optimizer in the convolutional neural network model is Adam function, and the learning rate is 0.0001.
In the present invention, the convolutional neural network model performs pixel analysis on the sub-picture obtained in step S2, and identifies regions with different morphological characteristics, such as cracks, scratches, and pores, by using an edge detection method.
According to the present invention, preferably, in step S4, the formula is:
CR＝m/M；
wherein, CR represents the crack rate of each original metallographic picture;
m represents the number of pixel points of the crack part of each original metallographic picture;
m represents the total pixel number of each original metallographic picture;
n represents the number of original metallographic pictures;
The technical scheme of the invention has the following beneficial effects:
the invention relates to a crack statistical method based on machine learning and aiming at high-temperature alloy additive manufacturing. According to the method, the picture file automatic preprocessing program is established through the os library and the PIL library in python to automatically process the original metallographic picture, the cracks are identified and counted by utilizing the Unet model and the convolutional neural network model of the semantic segmentation algorithm, and the crack rate of the high-temperature alloy sample manufactured by the additive materials can be rapidly calculated in batches. Compared with the existing manual binary segmentation method and automatic binary segmentation method, the method has the advantages that when the machine learning algorithm is applied to carry out crack rate statistics, the statistical result is accurate, the statistical task can be completed rapidly and massively, and manual intervention and reprocessing are not needed.
Additional features and advantages of the invention will be set forth in the detailed description which follows.
Drawings
The above and other objects, features and advantages of the present invention will become more apparent by describing in detail exemplary embodiments thereof with reference to the attached drawings, wherein like reference numerals generally represent like parts in the exemplary embodiments of the present invention.
Fig. 1 shows a schematic process flow diagram of a statistical method for additive manufacturing cracks in a superalloy provided by an embodiment of the present invention.
FIG. 2 shows a macro topography of an additive manufactured superalloy sample prepared by a method provided by an embodiment of the present invention after corundum blasting.
Fig. 3 shows a flow chart of superalloy additive manufacturing and cutting a cross-section provided by an embodiment of the present invention.
Fig. 4(a) - (c) show original metallographic pictures of cross sections cut from additive manufactured superalloy samples provided by embodiments of the present invention.
Fig. 5 shows a sub-picture of an original metallographic picture of an additive manufactured superalloy sample processed by an automatic picture file preprocessing procedure according to an embodiment of the present invention.
Detailed Description
Preferred embodiments of the present invention will be described in more detail below. While the following describes preferred embodiments of the present invention, it should be understood that the present invention may be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
Example 1
The embodiment provides a statistical method for a superalloy additive manufacturing crack, as shown in fig. 1, the statistical method includes the following steps:
s1: selecting a GH3536 nickel-based high-temperature alloy base material, and carrying out heat treatment on the base material in an as-cast state, wherein the size of the base material is 20mm multiplied by 20 mm; polishing a base material by adopting 240-mesh sand paper, performing corundum sand blasting treatment, and treating the base material by using an argon atomization method to obtain DZ98M high-temperature alloy powder (the granularity is 80-250 meshes); the method comprises the following steps of performing laser additive manufacturing on DZ98M high-temperature alloy powder by using a laser coaxial powder feeding additive manufacturing system to obtain an additive manufacturing high-temperature alloy sample, wherein the specific process parameters are as follows: the laser power is 1800W, the diameter of a laser spot is 1.0mm, the scanning speed is 1300mm/min, the powder feeding speed is 11.0g/min, the powder feeding gas is argon, and the pressure of the protective gas is 0.16 MPa. The macroscopic morphology of the additive manufactured superalloy sample subjected to corundum sand blasting is shown in fig. 2.
Cutting the additive manufacturing superalloy sample according to the position shown in FIG. 3, dividing the sample into four parts by warp cutting, grinding three sections (A, B, C) of the cut sample by 150-mesh, 400-mesh, 1000-mesh and 2000-mesh sand paper respectively, and polishing by using W2.5 polishing paste;
shooting the whole area of each of the three polished sections by using a metallographic microscope to obtain original metallographic pictures corresponding to the three sections, as shown in fig. 4(a) - (c);
s2: an os library and a PIL library function based on python are adopted to establish an automatic picture file preprocessing program; cutting each original metallographic picture into a plurality of sub-pictures with the same size (128 pixels by 128 pixels) by using the picture file automatic preprocessing program, as shown in fig. 5;
s3: writing an Unet model and a convolutional neural network model of a semantic segmentation algorithm through a keras library function of python; segmenting the sub-picture obtained in the step S2 by using the Unet model of the semantic segmentation algorithm, determining cracks, scratches and pore parts (namely identifying crack regions, converting the crack regions into red color phases and distinguishing the red color phases from other regions), and automatically identifying the cracks, the scratches and the pores by using the convolutional neural network model;
s4: and counting the pixel points of the crack part and the total pixel points of each original metallographic picture, calculating the crack rate of each original metallographic picture by using a formula, and then calculating the crack rate of the additive manufacturing high-temperature alloy sample.
The formula is:
CR＝m/M；
wherein, CR represents the crack rate of each original metallographic picture;
m represents the pixel point number of the crack part of each original metallographic picture;
m represents the total pixel point number of each original metallographic picture;
n represents the number of original metallographic pictures;
Calculating the crack rates CR of the sections corresponding to each original metallographic picture to be 0.3603%, 0.6307% and 0.2991% respectively, and calculating the average value to obtain the average crack rate of the sample
(i.e., the crack rate of the additive manufactured superalloy sample) was 0.4300%.
The results of the above embodiments show that the method of the present invention adopts python language to program, can automatically and rapidly count the cracks of the high temperature alloy sample for additive manufacturing, can be used as a computer-assisted section for evaluating the crack sensitivity of the high temperature alloy for additive manufacturing, greatly reduces the man-hour of manual statistics, and can also be used for designing the high temperature alloy for additive manufacturing without cracks and with high performance.
While embodiments of the present invention have been described above, the above description is illustrative, not exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments.
Claims (10)
1. A statistical method for high-temperature alloy additive manufacturing cracks is characterized by comprising the following steps:
s1: performing laser additive manufacturing on the high-temperature alloy powder by using an additive manufacturing system to obtain an additive manufactured high-temperature alloy sample; cutting the material additive manufacturing high-temperature alloy sample, and grinding and polishing to obtain a plurality of sections; shooting the whole area of each section after polishing by using a metallographic microscope to obtain a corresponding original metallographic picture;
s2: establishing an automatic preprocessing program of the picture file; cutting each original metallographic picture into a plurality of sub-pictures with the same size by using the picture file automatic preprocessing program;
s3: writing a Unet model and a convolutional neural network model of a semantic segmentation algorithm; performing semantic segmentation on the sub-picture obtained in the step S2 by using the Unet model of the semantic segmentation algorithm, determining cracks, scratches and pore parts, and automatically identifying the cracks, the scratches and the pores by using the convolutional neural network model;
s4: and counting the pixel points of the crack part and the total pixel points of each original metallographic picture, calculating the crack rate of each original metallographic picture by using a formula, and then calculating the crack rate of the additive manufacturing high-temperature alloy sample.
2. The statistical method of superalloy additive manufacturing cracks of claim 1, wherein, in step S1,
the additive manufacturing system is a laser coaxial powder feeding additive manufacturing system or a laser powder bed additive manufacturing system;
the high-temperature alloy powder is prepared by carrying out an argon atomization method on a high-temperature alloy base material, and the granularity of the high-temperature alloy powder is 80-250 meshes.
3. The statistical method for superalloy additive manufacturing cracks of claim 1, wherein in step S1, the cutting is wire cutting the additive manufacturing superalloy sample.
4. The statistical method for superalloy additive manufacturing cracks of claim 1, wherein in step S1, the magnification of the metallographic microscope is selected to be 100 times or more during the photographing;
the original metallographic picture adopts an RGB mode.
5. The statistical method of superalloy additive manufacturing cracks of claim 1, wherein, in step S2,
adopting an os library and a PIL library function based on python to establish the automatic picture file preprocessing program;
the automatic picture file preprocessing program comprises picture file loading, picture file cutting, picture renaming and picture file modular storage.
6. The statistical method of superalloy additive manufacturing cracks of claim 5, wherein the sub-picture is 64 pixels by 64 pixels, 128 pixels by 128 pixels, or 256 pixels by 256 pixels in size.
7. The statistical method for superalloy additive manufacturing cracks of claim 1, wherein in step S3, the Unet model and the convolutional neural network model of the semantic segmentation algorithm are each written independently by a keras library function of python.
8. The statistical method of superalloy additive manufacturing cracks of claim 7, wherein the convolutional neural network model comprises a convolutional layer, a pooling layer, and a fully connected layer.
9. The statistical method of superalloy additive manufacturing cracks of claim 7, wherein an optimizer in the convolutional neural network model is an Adam function with a learning rate of 0.0001.
10. The superalloy additive manufacturing crack statistical method of claim 1,
in step S4, the formula is:
CR＝m/M；
wherein, CR represents the crack rate of each original metallographic picture;
m represents the number of pixel points of the crack part of each original metallographic picture;
m represents the total pixel point number of each original metallographic picture;
n represents the number of original metallographic pictures;
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210258918.0A CN114782313A (en) | 2022-03-16 | 2022-03-16 | Statistical method for high-temperature alloy additive manufacturing cracks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210258918.0A CN114782313A (en) | 2022-03-16 | 2022-03-16 | Statistical method for high-temperature alloy additive manufacturing cracks |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114782313A true CN114782313A (en) | 2022-07-22 |
Family
ID=82425401
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210258918.0A Pending CN114782313A (en) | 2022-03-16 | 2022-03-16 | Statistical method for high-temperature alloy additive manufacturing cracks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114782313A (en) |
-
2022
- 2022-03-16 CN CN202210258918.0A patent/CN114782313A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110579473B (en) | Automatic full-field quantitative statistical distribution characterization method for dendritic crystal structure in metal material | |
| CN111696632B (en) | Method for characterizing full-view-field quantitative statistical distribution of gamma' -phase microstructure in metal material | |
| US7602963B2 (en) | Method and apparatus for finding anomalies in finished parts and/or assemblies | |
| US20170372470A1 (en) | Method of separating, identifying and characterizing cracks in 3d space | |
| US10106876B2 (en) | Method of surface-treating a cast intermetallic component | |
| CN104897698A (en) | Method for three-dimensional representation of stratified structure micrometer CT imaging of turbine blade heat barrier coating | |
| CN108489986A (en) | A kind of increasing material manufacturing on-line checking and restorative procedure | |
| CN110992318A (en) | Special metal flaw detection system based on deep learning | |
| CN113781458A (en) | Artificial intelligence based identification method | |
| JPWO2018020681A1 (en) | Setting method, inspection method, defect evaluation apparatus and manufacturing method of structure | |
| Kladovasilakis et al. | A case study on 3D scanning, digital reparation and rapid metal additive manufacturing of a centrifugal impeller | |
| WO2018228251A1 (en) | Method for removing cracks on inner cavity surface of part formed by selective laser melting | |
| CN111738131A (en) | Method for extracting parameter characteristics of alloy two-phase microstructure | |
| CN114782313A (en) | Statistical method for high-temperature alloy additive manufacturing cracks | |
| EP3536419A1 (en) | Surface or interface defect detection | |
| CN113351881B (en) | Mixed additive manufacturing method for aero-engine case | |
| CN112507929A (en) | Vehicle body spot welding slag accurate detection method based on improved YOLOv3 network | |
| CN109115575B (en) | Method for detecting inclusions in metal 3D printing nickel-based superalloy | |
| CN116223214A (en) | Evaluation method for mechanical properties of GH4169 alloy thin-wall part manufactured by additive | |
| CN107255450B (en) | Screening method of tungsten carbide alloy powder | |
| CN115170484A (en) | Characterization and classification method for surface defects of laser additive manufacturing workpiece | |
| Bai et al. | Research on surface crack detection based on computer image recognition | |
| Lin et al. | Alloy Cast Product Defect Detection Based on Object Detection | |
| CN117451727B (en) | Quality control method for nozzle machining process | |
| CN113469988A (en) | Defect identification method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |