CN117635543A - Forward tracking and fault reverse positioning method and system for internal defects of metal - Google Patents
Forward tracking and fault reverse positioning method and system for internal defects of metal Download PDFInfo
- Publication number
- CN117635543A CN117635543A CN202311488555.0A CN202311488555A CN117635543A CN 117635543 A CN117635543 A CN 117635543A CN 202311488555 A CN202311488555 A CN 202311488555A CN 117635543 A CN117635543 A CN 117635543A
- Authority
- CN
- China
- Prior art keywords
- defects
- defect
- tracked
- candidate
- image
- 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.)
- Granted
Links
- 230000007547 defect Effects 0.000 title claims abstract description 370
- 238000000034 method Methods 0.000 title claims abstract description 68
- 230000002441 reversible effect Effects 0.000 title claims abstract description 31
- 239000002184 metal Substances 0.000 title claims abstract description 17
- 229910052751 metal Inorganic materials 0.000 title claims abstract description 17
- 239000000463 material Substances 0.000 claims abstract description 61
- 238000013519 translation Methods 0.000 claims abstract description 29
- 230000011218 segmentation Effects 0.000 claims abstract description 17
- 239000007769 metal material Substances 0.000 claims abstract description 13
- 239000000203 mixture Substances 0.000 claims abstract description 12
- 230000001052 transient effect Effects 0.000 claims abstract description 10
- 230000007613 environmental effect Effects 0.000 claims abstract description 9
- 238000013528 artificial neural network Methods 0.000 claims abstract description 7
- 239000013598 vector Substances 0.000 claims description 18
- 239000011159 matrix material Substances 0.000 claims description 17
- 238000004422 calculation algorithm Methods 0.000 claims description 13
- 238000011065 in-situ storage Methods 0.000 claims description 12
- 238000003860 storage Methods 0.000 claims description 10
- 238000002591 computed tomography Methods 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 238000003709 image segmentation Methods 0.000 claims description 6
- 230000004807 localization Effects 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 5
- 230000006870 function Effects 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 3
- 230000005489 elastic deformation Effects 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000010845 search algorithm Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 description 19
- 238000004458 analytical method Methods 0.000 description 9
- 238000011038 discontinuous diafiltration by volume reduction Methods 0.000 description 5
- 230000007246 mechanism Effects 0.000 description 5
- 238000001514 detection method Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000002950 deficient Effects 0.000 description 2
- 230000001066 destructive effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 229910000831 Steel Inorganic materials 0.000 description 1
- 238000002441 X-ray diffraction Methods 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000001493 electron microscopy Methods 0.000 description 1
- 238000012613 in situ experiment Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000000873 masking effect Effects 0.000 description 1
- 150000002739 metals Chemical class 0.000 description 1
- 230000006911 nucleation Effects 0.000 description 1
- 238000010899 nucleation Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000006116 polymerization reaction Methods 0.000 description 1
- 239000010959 steel Substances 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000002604 ultrasonography Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
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
-
- 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
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural 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
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- 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]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Analysing Materials By The Use Of Radiation (AREA)
Abstract
The invention discloses a method and a system for forward tracking and reverse fault positioning of internal defects of metal, wherein the method comprises the following steps: acquiring two-dimensional tomographic images of a plurality of transient states inside a metal material sample; dividing a defect from the two-dimensional tomographic image based on a neural network and threshold segmentation, generating a binary image containing only the defect and a background; based on a three-dimensional space clustering method, representing defects in the binary image as point clouds; identifying all possible target candidate defects of the tracked defects in a set range of adjacent states by rotating and translating the tracked defects, and performing the same rotation and translation on the environment information; calculating the similarity between the tracked defect and the target candidate defect environmental point cloud, and further determining the optimal candidate defect; reversely establishing defect compositions of the final fracture of the material in each stretching state, and determining the original key defects of final failure; the internal defects of the material can be extracted, and the edge defects and the fracture defects can be extracted at the same time, so that the accuracy of identifying the defects is improved.
Description
Technical Field
The invention belongs to the technical field of intelligent analysis of material defects, and particularly relates to a method and a system for forward tracking and reverse fault positioning of internal defects of metals.
Background
Material failure carries significant risks in various industries, and remains a major research focus in materials science. These failures often originate from the rapid evolution of defects and are affected by factors such as mechanical stress, temperature changes, and chemical interactions. With the advent of detection techniques such as X-ray diffraction, electron microscopy, and ultrasound, researchers have been given the ability to identify such defects on multiple scales. Among them, X-ray computed tomography (XCT) is widely used in its non-destructive method and in the clear visualization of the internal structure of materials. When combined with in situ experiments, it provides a dynamic view of the process of gradual changes in the internal structure of the material. This unique capability makes in situ XCT a very important tool for studying complex mechanisms of material failure.
In recent years, researchers at home and abroad use in-situ XCT to widely apply the material failure problem. For example, by studying the relationship of defects to strain and triaxial stresses, to enhance the understanding of high strength steel damage; by quantifying nucleation, growth and polymerization of defects, the ductile failure mechanism of SA508 Gr.3 is deeply understood; edge failure of hinge and shear holes is predicted by a micromechanically driven model. Most of these studies use global statistical methods to analyze defect evolution, where the metrics involved include average porosity, global deformation, and the amount of change in the representative defect set, among others. Such statistics may not be sufficient to adequately capture detailed evolution rules of individual defects, and the rapid evolution of a single defect may have a critical impact on material failure. Thus, a finer field of view is required. Specifically, it is to track individual defects from their initial state to their failure state and establish their evolving correlation throughout the life cycle. This correlation can also help us identify the original critical defect that caused the eventual failure of the material.
The research results on defect tracking algorithms in materials science are limited. Of these, the pioneering work of the foreign scholars lecame, who used a graph-based approach to track defects under tensile load, revealed mechanisms behind heterogeneous cavity growth. In this algorithm, a graph is used to represent the relationship between all defects, where each node represents a single defect and each edge represents the cost of the connection between two nodes, which is defined as the core of the algorithm, by the following equation:
wherein, c i -c j Measuring Euclidean distance, t, between two defects j ＝t i +1 limits the moment in time, |v i ≤v j |≤τ vol Representing the tolerance of adjacent state volume changes.
The algorithm relies primarily on small displacements and volume changes of the defect to establish the correlation. However, it still faces many challenges such as XCT resolution, defect segmentation accuracy, and defect aggregation. In addition, this algorithm also ignores changes in spatial morphology and invariance in the local environment of the defect. This is an important feature of defects in the adjacent state, meaning that the morphology of the defects may change significantly during stretching, but their environment in the localized area (i.e., the defects surrounding the defect) remains relatively similar.
In situ X-ray computed tomography (XCT) provides a non-destructive method for studying the dynamic behavior of defects inside a material. There have been a great deal of research using it to explore the failure mechanism of metallic materials, but most of these have focused on global statistics on defect data. Such global statistics often fail to capture the detailed evolution of individual defects, while some studies have shown that the rapid evolution of an individual defect inside a material can have catastrophic consequences for the failure of the material. Therefore, in the in-situ XCT experiment process, it is very important to establish the association relationship between defects and to study individual features of defect evolution.
Defect segmentation is a precondition for analysis, whereas conventional methods (such as chinese patent publication No. CN113298757 a) focus on the detection of "internal defects", and ignore the identification of "edge defects" before fracture and "fracture defects" after fracture, and thus cannot acquire the complete failure process of the material. When the failure mechanism of the metal material is researched, most of the traditional methods stay on the global statistical method, and cannot effectively capture the independent evolution process of all individual defects, so that deeper analysis cannot be performed. Although the conventional in-situ XCT can obtain all defects of each transient state in the material stretching process, the association relationship of the defects in the time dimension cannot be established. In-situ XCT-based defect tracking methods in the field of material science are few, and only the defect positions and sizes are matched, so that the shape and the environmental characteristics of the defects are not considered, and the accuracy is limited. The effective method has not been found to be capable of establishing a reverse path for the final material failure, nor is it capable of reverse locating the original critical defect that caused the failure.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a forward tracking and fault reverse positioning method for internal defects of metal, which automatically tracks all independent defects in a material from an initial state to a failure state and establishes correlation of the defects in a continuous stretching state. The method represents the defect as a three-dimensional point cloud, and fully utilizes the visual characteristics of the three-dimensional point cloud, including the spatial morphology and local environment information of the defect. Compared with the traditional method, the method improves the average tracking accuracy from 86.1% to 92.7%. In addition, these tracking results can be used to reconstruct the failure path in reverse and locate the original critical defects that lead to the final failure, and the present invention has been successfully applied to the analysis of the evolution of defects in aviation materials and has proven to be practical.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: a method for forward tracking and reverse positioning of defects in metal comprises the following steps:
acquiring two-dimensional tomographic images of a plurality of transient states inside a metal material sample;
dividing a defect from the two-dimensional tomographic image based on a neural network and threshold segmentation, generating a binary image containing only the defect and a background;
using a three-dimensional space clustering method to represent the defects in the binary image as three-dimensional point clouds;
identifying all possible target candidate defects of the tracked defects in a set range of adjacent states by rotating and translating the tracked defects based on the defects represented by the three-dimensional point cloud, then utilizing the environment information of the source object to perform the same rotation and translation on the environment information, obtaining a target candidate defect set of each tracked defect, and calculating an optimal rotation matrix and a translation vector corresponding to each candidate defect;
calculating the similarity between the tracked defect and the target candidate defect environmental point cloud, and determining the optimal candidate defect according to the similarity;
and reversely establishing the defect composition of the final fracture of the material in each tensile state in a tree topology mode according to the target candidate defect set, the optimal rotation matrix and translation vector corresponding to each candidate defect and the optimal candidate defects, and reversely determining the final invalid original key defects.
Further, the two-dimensional tomographic images of the plurality of transient states inside the metal material sample include: during in situ stretching, 7 different phases of tomographic images of the specimen were captured by XCT, including an initial state, 2 elastic deformation states, 2 plastic deformation states, a pre-fracture state, and a post-fracture state, with the different phases of states corresponding to strains of 0% (initial state), 9.6%, 25%, 40%, 50.8%, 58.2%, and 63.8%, respectively.
Further, based on the three-dimensional spatial clustering method, the representing the defect in the binary image as a point cloud includes:
defects are classified into 3 types according to their stress conditions and spatial locations: internal defects, edge defects and fracture defects;
the method comprises the steps of extracting defects in a mode of combining U-Net with threshold segmentation, specifically, a training set of the U-Net is an original CT scanning image, a tag set is a complete outline mask of material entities in the image, an encoder and a decoder in the U-Net are composed of two groups of 3X 3 convolution layers, a ReLU activation function is adopted, downsampling and upsampling are carried out through 2X 2 max pooling and 2X 2 transpose convolution, meanwhile, batch Normalization layers are added before the ReLU to enable training to be more stable, after the complete entity mask of the materials is obtained, the complete entity of the materials containing three types of defects is separated from a complex background of the original image, the difference between the intensities of the defects and the pixels of the materials in the image is obvious, and the intensities of the defects and the pixels of the materials in the image are segmented through an OSTU algorithm, so that a binary image containing only three types of defects and the background is formed.
Further, based on the three-dimensional spatial clustering method, the representing the defect in the binary image as a point cloud includes: in one slice, the defect is considered as a set of pixels, each of which is represented by its coordinates (x i ,y i ) Representing that each pixel is associated with a slice number z i Constructing a set of space coordinates, obtaining the maximum distance between two adjacent voxels of each single defect setting and the minimum voxel number in a defect cluster by using a space clustering algorithm DBSCAN, identifying noise points at the same time, and assigning a unique ID to each defect according to the number of space body pixel points in the defect.
Further, the following operation is performed for each tracked defect to obtain a candidate defect set of the tracked defects:
searching possible target defects in a spherical region with a set radius in the next state by utilizing the centroid of the tracked defect in the last state, performing volume filtering on all defects in the spherical region, wherein defects which do not meet the condition in the sphere range are noise defects, and the defects which meet the condition are candidate defects, and aligning the tracked defect and each candidate defect in the same coordinate system to obtain each candidate defect and an optimal rotation matrix R corresponding to each candidate defect i And translation vector t i 。
Further, calculating the similarity with the target candidate defect environment point cloud, and determining the optimal candidate defect according to the similarity comprises:
optimal rotation matrix R using target candidate defects i And translation vector t i Rotating and translating the environmental point cloud with the tracked defects to obtain a converted environmental point cloud;
and calculating the similarity between the converted ambient point cloud and the ambient point cloud of the target candidate defect to determine the best matching candidate.
Further, in the form of tree topology, reversely establishing the defect composition of the final fracture of the material in each tensile state, and reversely determining the original key defect of the final failure comprises:
adopting a depth-first search algorithm to reversely construct a reverse topology of a maximum fracture in a failure state, wherein the reverse topology is in a tree structure, a root node represents a fracture defect, each node is represented by a defect ID, and the connection between the nodes represents the composition relation;
all sub-defect nodes forming the maximum defect of the current state are dangerous defects, the maximum defect in each state is reversely tracked along the dangerous defects until the leaf nodes, and the original key defects causing material failure are identified.
The invention also provides a metal internal defect forward tracking and fault reverse positioning system which comprises an image acquisition module, an image segmentation module, an image conversion module, a transformation module, an optimal candidate defect acquisition module and a failure defect determination module;
the image acquisition module is used for acquiring two-dimensional tomographic images of a plurality of transient states in the metal material sample;
the image segmentation module segments defects from the two-dimensional tomographic image based on a neural network and threshold segmentation, and generates a binary image only containing the defects and the background;
the image conversion module uses a three-dimensional space clustering method to represent defects in the binary image as three-dimensional point clouds;
the transformation module is used for identifying all possible target candidate defects of the tracked defects in a set range of adjacent states based on the defects represented by the three-dimensional point cloud through rotation and translation, carrying out the same rotation and translation on the environment information of the source object, obtaining a target candidate defect set of each tracked defect, and calculating an optimal rotation matrix and a translation vector corresponding to each candidate defect;
the optimal candidate defect acquisition module is used for determining an optimal candidate defect according to the similarity by calculating the similarity of the tracked defect and the target candidate defect environment point cloud;
and the failure defect determining module reversely establishes the defect composition of the final fracture of the material in each tensile state in a tree topology mode according to the target candidate defect set, the optimal rotation matrix and translation vector corresponding to each candidate defect and the optimal candidate defect, and reversely determines the original key defect which is finally failed.
The invention also provides a computer device, which comprises a processor and a memory, wherein the memory is used for storing a computer executable program, the processor reads the computer executable program from the memory and executes the computer executable program, and the processor can realize the forward tracking and fault reverse positioning method of the metal internal defects when executing the computer executable program.
A computer readable storage medium having a computer program stored therein, which when executed by a processor, implements the method for forward tracing and reverse fault localization of metal internal defects according to the present invention.
Compared with the prior art, the invention has at least the following beneficial effects:
the invention focuses on the dynamic evolution rule of the defects in the whole period from the original state of the material to the tensile fracture state of the material, but not the defects in a specific state; according to the invention, the defect information of the material in a plurality of instantaneous states is obtained through the in-situ XCT, and the method is different from the prior method, so that the internal defects of the material are extracted, the edge defects and the fracture defects are extracted at the same time, and the whole life cycle from the initial stage to the failure of the material in the stretching process is covered;
the defect tracking is performed based on the defect form and the environment information thereof, and experiments prove that the accuracy is improved from 86.1% to 92.7% compared with the traditional method;
the invention establishes the association relation among all independent defects, so that the analysis of the defects is not remained on a global analysis method, and the analysis is not carried out only by relying on global, average or partial statistical data, but is focused on the evolution analysis of the independent defects more deeply, and basic data is provided for further analysis modeling;
the invention can more intuitively analyze the evolution rule of the defects by establishing the reverse topology of the defects, identify the most original defect source and provide a basis for the reliability detection and the risk early warning of the materials.
Drawings
FIG. 1 is an overall flow chart of a method provided by the present invention.
FIG. 2 illustrates binary segmentation of an exemplary CT image in accordance with the present invention.
FIG. 3 is a three-dimensional representation of an example defect.
Fig. 4 is a defect tracking algorithm-searching for possible candidates.
Fig. 5 is a defect tracking algorithm-calculating the best candidate.
Fig. 6 is an evolving topology of defects.
FIG. 7 is a schematic diagram showing the defect evolution process along with the stretching process.
Fig. 8 is a schematic diagram of the evolution process of the original crack source.
Detailed Description
The invention establishes the association relation in the stretching process of the individual defects in the material based on the in-situ XCT, and reversely locates the original key defects which cause the final failure of the material by establishing the evolution topology. The method for forward tracking and reverse positioning of the defects in the metal, provided by the invention, comprises the following steps of:
stretch scanning: and carrying out in-situ stretching and CT scanning on the metal material sample to obtain two-dimensional tomographic images of a plurality of transient states in the metal material sample.
Secondly, defect segmentation: the defects are separated from the image based on a neural network and a threshold segmentation technique, and a binary image containing only the defects and the background is generated.
Three-dimensional representation: based on a three-dimensional spatial clustering algorithm, the defects in the binary image are expressed in the form of point clouds so as to utilize the spatial characteristics of the points.
Fourth, defect tracking: by rotating and translating the tracked defects, identifying all possible target candidate defects in the set range of adjacent states, and then utilizing the environment information (including the defects and the peripheral defects) of the source object, carrying out the same rotation and translation on the defects, and calculating the similarity with the environment of the target candidate to determine the optimal target object.
Fifth, reverse topology: and reversely establishing the defect composition of the final fracture of the material in each stretching state in a tree topology form, and reversely determining the original key defect of final failure.
The method is described in detail in connection with examples as follows:
s1: in the process of in-situ stretching, tomographic images of 7 different stages of the specimen including an initial state, 2 elastic deformation states, 2 plastic deformation states, a pre-fracture state, a post-fracture state are captured by XCT. The strains corresponding to these states were 0% (initial state), 9.6%, 25%, 40%, 50.8%, 58.2% and 63.8% (failure state), respectively, covering the entire life cycle of the metallic material. For each particular state, approximately 990 tomographic images were collected, each image having a resolution of 1013 x 998 pixels, with both pixel size and layer thickness of 1.35 microns. Finally, a 1013×998×990×7 4-dimensional dataset is obtained.
S2: defects are classified into 3 types according to their stress conditions and spatial locations: internal defects, edge defects and fracture defects. Since edge defects and fracture defects can cause defective portions to merge with the scanned background, extracting this type of defect requires first predictively restoring the full profile of the material entity. The invention adopts a mode of combining U-Net and threshold segmentation technology to extract defects. The training set of the U-Net is an original CT scanning image, and the tag set is a complete contour mask of material entities in the image. The complete entity of the material containing the three types of defects can be separated from the background by masking, at which time the pixel intensities between the background, the entity of the material and the defects have a significant difference, and the defects can be extracted directly by using the threshold segmentation technique OSTU. The final result is a binary image containing three types of defects and background. The above process is shown in fig. 2. The training set of the U-Net is an original CT scanning image, the tag set is a complete outline mask of material entities in the image, the encoder and decoder parts in the U-Net are composed of two groups of 3×3 convolution layers, a ReLU activation function is adopted, downsampling and upsampling are carried out through 2×2 max pooling and 2×2 transposed convolution, meanwhile, batch Normalization layers are added before the ReLU to enable training to be more stable, after the material complete entity mask is obtained, the material complete entity containing three types of defects is separated from the complex background of the original image, the difference between the defect and the material pixel intensity in the image is obvious, and the defect and the material pixel intensity in the image are divided through an OSTU algorithm to form a binary image containing only three types of defects and the background.
S3: after obtaining the defects of the segmentation in the two-dimensional image, they need to be represented in three-dimensional form for use in the subsequent tracking algorithm. A single defect is typically present in multiple successive slices, in which the defect can be seen as a set of pixels, each pixel being defined by its coordinates (x i ,y i ) And (3) representing. By associating one slice number z per pixel i A set of spatial coordinates may be constructed. Finally, each single defect is obtained by using a spatial clustering algorithm DBSCAN, the maximum distance between two adjacent voxels is set to be 3, the minimum voxel number in one defect cluster is set to be 5, and the defect clusters consisting of less than 5 points are regarded as noise. And assigning a unique ID to each defect according to the number of the space voxel points, and marking the association relation of the defects in the subsequent tracking process. Above is passed throughCheng Ru is shown in fig. 3.
S4: with the centroid of the tracked defect in the last state, the possible target defect is searched for in the spherical region in the next state, and as a preferred example, the tracking effect is the best when the spherical region radius is set to 150 voxel distance. In consideration of the influence of factors such as CT resolution and segmentation accuracy, the volume reduction of a single defect in an adjacent stretching state cannot be more than 20%, and 20% is the tolerance of the volume reduction. Thus, volumetric filtering of all defects within the spherical region may reduce subsequent computational redundancy. In this process, defects within the sphere that do not meet the condition are referred to as "noise defects", and defects that meet the condition are referred to as "candidate defects"; the tolerance for volume reduction was set to 20%. If the volume reduction is greater than 20% is considered a noise defect, a volume reduction of less than 20% or a volume increase is a target candidate defect. Further, the alignment operation of the tracked defect and each candidate object thereof is performed in the same coordinate system, and the specific definition is as follows:
(1) Point cloud X of tracked defect and point cloud C of ith candidate defect i The definition is as follows:
X＝{x 1 ,x 2 ,…,x n } x j ∈R 3
C i ＝{c i1 ,c i2 ,…,c in } c ij ∈R 3
wherein x is j And c ij The coordinates of the j-th point in the respective defect point cloud.
(2) By minimizing the distance between matching points, X and C are achieved i The process can be expressed as:
wherein N is x Represents the total number of points in the tracking defect, R i And t i The best rotation matrix and translation vector, respectively.
The above process is as shown in FIG. 4As shown, for each tracked defect, a target candidate defect set thereof is obtained, and each candidate defect calculates an optimal rotation matrix R corresponding thereto, via S4 i And translation vector t i 。
S5: and determining the best candidate defect by calculating the similarity of the environmental point clouds between the tracked defect and the target candidate defect. The ambient point cloud is defined as the defect itself and other defects within a sphere around it, the sphere radius is set to 100 voxel distance, based on the concept that rotation and translation of the defect in a certain local area exhibit similar patterns, thus sharing similar rotation matrices and translation vectors, and the detailed calculation process is defined as follows:
(1) Ambient point cloud of tracked defect X 'and ith candidate defect C' i The ambient point cloud of (2) is as follows:
X′＝{x′ 1 ,x′ 2 ,…,x′ n } x′ j ∈R 3
C′ i ＝{c′ i1 ,c′ i2 ,…,c′ in } c′ ij ∈R 3
wherein x' j And c' ij The coordinates of the j-th point in the respective defective ambient point cloud.
(2) Utilizing the candidate defect optimal rotation matrix R obtained in the last step i And translation vector t i To rotate and translate the ambient point cloud of the tracked defect to obtain a converted ambient point cloud X':
X″＝R i *X′+t i
(3) Calculation of X 'and C' i Similarity S of (2) i To determine the best match candidate:
wherein N represents the number of points in X',calculated are X 'and C' i Is the minimum distance between the average points of (D) i SAR represents the centroid distance of the tracked defect and the candidate defect, and SAR is the size of the search area>For limiting the distance of movement of the candidate defects, the final similarity S i Fall to [0,1 ]]Within the range, a larger value indicates a higher degree of similarity.
As shown in fig. 5, by introducing the ambient point cloud similarity, the best candidate defect can be calculated from the candidate set, which has not only the similar defect itself, but also a similar environment.
S6: based on the forward evolution relation between defects in each state obtained in S4 and S5, the forward tracking result of the depth-first search algorithm S4 and S5 is stored in a JSON file in the form of a key value pair of PreStat-PreDefectId: nextStat-NextDefectId. Traversing the JSON file to complete fault topology creation and reverse positioning through a depth-first search (DFS) strategy; the method starts (break defect) processing from the root node by a recursive function, adds the current node to the topology tree, and calls the recursive function to process each ordered child node until the current node has no leaf nodes. Reverse topology of the maximum fracture in the failure state is constructed reversely, the reverse topology is in a tree structure, wherein a root node represents a fracture defect, each node is represented by a defect ID, and the connection between the nodes represents the composition relation.
As shown in fig. 6, as an example, all the sub-defect nodes constituting the maximum defect of the current state in the tree structure may be defined as "dangerous defects", which are key factors causing material failure. Other defects in the tree structure are related to material failure but are not a major factor and are therefore defined as "indeterminate defects". Defects not shown in the tree structure have no relation to the final failure and are therefore defined as "security defects".
Finally, the biggest defect in each state is traced back along the "dangerous defect" until the leaf node, which is a defect with id=2 in the S2 state in fig. 6 as an example, and the defect evolution process can refer to fig. 7 and 8, and the original key defect causing the material failure can be successfully identified.
The invention also provides a metal internal defect forward tracking and fault reverse positioning system which comprises an image acquisition module, an image segmentation module, an image conversion module, a transformation module, an optimal candidate defect acquisition module and a failure defect determination module;
the image acquisition module is used for acquiring two-dimensional tomographic images of a plurality of transient states in the metal material sample;
the image segmentation module segments defects from the two-dimensional tomographic image based on a neural network and threshold segmentation, and generates a binary image only containing the defects and the background;
the image conversion module uses a three-dimensional space clustering method to represent defects in the binary image as point clouds;
the transformation module is used for identifying all possible target candidate defects of the tracked defects in a set range of adjacent states based on the defects represented by the three-dimensional point cloud through rotation and translation, carrying out the same rotation and translation on the environment information of the source object, obtaining a target candidate defect set of each tracked defect, and calculating an optimal rotation matrix and a translation vector corresponding to each candidate defect;
the optimal candidate defect acquisition module is used for determining an optimal candidate defect according to the similarity by calculating the similarity of the tracked defect and the target candidate defect environment point cloud;
and the failure defect determining module reversely establishes the defect composition of the final fracture of the material in each tensile state in a tree topology mode according to the target candidate defect set, the optimal rotation matrix and translation vector corresponding to each candidate defect and the optimal candidate defect, and reversely determines the original key defect which is finally failed.
In another aspect, the present invention provides a computer readable storage medium, where a computer program is stored, where the computer program, when executed by a processor, can implement the method for forward tracking and reverse fault localization of internal defects of a metal according to the present invention.
The computer device may be a notebook computer, a desktop computer, or a workstation.
The processor may be a Central Processing Unit (CPU), a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), or an off-the-shelf programmable gate array (FPGA).
The memory can be an internal memory unit of a notebook computer, a desktop computer or a workstation, such as a memory and a hard disk; external storage units such as removable hard disks, flash memory cards may also be used.
Computer readable storage media may include computer storage media and communication media. Computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. The computer readable storage medium may include: read Only Memory (ROM), random access Memory (RAM, random Access Memory), solid state disk (SSD, solid State Drives), or optical disk, etc. The random access memory may include resistive random access memory (ReRAM, resistance Random Access Memory) and dynamic random access memory (DRAM, dynamic Random Access Memory), among others.
Claims (10)
1. The method for forward tracking and reverse fault positioning of the metal internal defects is characterized by comprising the following steps of:
acquiring two-dimensional tomographic images of a plurality of transient states inside a metal material sample;
dividing a defect from the two-dimensional tomographic image based on a neural network and threshold segmentation, generating a binary image containing only the defect and a background;
using a three-dimensional space clustering method to represent the defects in the binary image as three-dimensional point clouds;
identifying all possible target candidate defects of the tracked defects in a set range of adjacent states by rotating and translating the tracked defects based on the defects represented by the three-dimensional point cloud, then utilizing the environment information of the source object to perform the same rotation and translation on the environment information, obtaining a target candidate defect set of each tracked defect, and calculating an optimal rotation matrix and a translation vector corresponding to each candidate defect;
calculating the similarity between the tracked defect and the target candidate defect environmental point cloud, and determining the optimal candidate defect according to the similarity;
and reversely establishing the defect composition of the final fracture of the material in each tensile state in a tree topology mode according to the target candidate defect set, the optimal rotation matrix and translation vector corresponding to each candidate defect and the optimal candidate defects, and reversely determining the final invalid original key defects.
2. The method of forward tracking and reverse fault localization of a metal internal defect according to claim 1, wherein the two-dimensional tomographic images of a plurality of transient states inside the metal material sample comprise: during in situ stretching, 7 different phases of tomographic images of the specimen were captured by XCT, including an initial state, 2 elastic deformation states, 2 plastic deformation states, a pre-fracture state, and a post-fracture state, with the different phases of states corresponding to strains of 0% (initial state), 9.6%, 25%, 40%, 50.8%, 58.2%, and 63.8%, respectively.
3. The method of claim 1, wherein representing defects in the binary image as point clouds based on a three-dimensional spatial clustering method comprises:
defects are classified into 3 types according to their stress conditions and spatial locations: internal defects, edge defects and fracture defects;
the method comprises the steps of extracting defects in a mode of combining U-Net with threshold segmentation, specifically, a training set of the U-Net is an original CT scanning image, a tag set is a complete outline mask of material entities in the image, an encoder and a decoder in the U-Net are composed of two groups of 3X 3 convolution layers, a ReLU activation function is adopted, downsampling and upsampling are carried out through 2X 2 max pooling and 2X 2 transpose convolution, meanwhile, batch Normalization layers are added before the ReLU to enable training to be more stable, after the complete entity mask of the materials is obtained, the complete entity of the materials containing three types of defects is separated from a complex background of the original image, the difference between the intensities of the defects and the pixels of the materials in the image is obvious, and the intensities of the defects and the pixels of the materials in the image are segmented through an OSTU algorithm, so that a binary image containing only three types of defects and the background is formed.
4. The method of claim 1, wherein representing defects in the binary image as point clouds based on a three-dimensional spatial clustering method comprises: in one slice, the defect is considered as a set of pixels, each of which is represented by its coordinates (x i ,y i ) Representing that each pixel is associated with a slice number z i Constructing a set of space coordinates, obtaining the maximum distance between two adjacent voxels of each single defect setting and the minimum voxel number in a defect cluster by using a space clustering algorithm DBSCAN, identifying noise points at the same time, and assigning a unique ID to each defect according to the number of space body pixel points in the defect.
5. The method of claim 1, wherein for each tracked defect, the following operations are performed to obtain the candidate defect set of tracked defects:
searching possible target defects in a spherical area with a set radius in the next state by utilizing the centroid of the tracked defects in the last state, performing volume filtering on all defects in the spherical area, wherein defects which do not meet the condition in the sphere range are noise defects, and the defects which meet the condition are candidate defects, wherein the tracked defects are identical to the candidate defectsAligning in each coordinate system to obtain each candidate defect and an optimal rotation matrix R corresponding to each candidate defect i And translation vector t i 。
6. The method of claim 1, wherein calculating a similarity to a target candidate defect ambient point cloud, and determining the best candidate defect based on the similarity comprises:
optimal rotation matrix R using target candidate defects i And translation vector t i Rotating and translating the environmental point cloud with the tracked defects to obtain a converted environmental point cloud;
and calculating the similarity between the converted ambient point cloud and the ambient point cloud of the target candidate defect to determine the best matching candidate.
7. The method of forward tracking and reverse fault localization of internal metal defects according to claim 1, wherein reverse establishing the composition of defects of the final fracture of the material in each tensile state in the form of a tree topology and reverse determining the original critical defects of the final failure comprises:
adopting a depth-first search algorithm to reversely construct a reverse topology of a maximum fracture in a failure state, wherein the reverse topology is in a tree structure, a root node represents a fracture defect, each node is represented by a defect ID, and the connection between the nodes represents the composition relation;
all sub-defect nodes forming the maximum defect of the current state are dangerous defects, the maximum defect in each state is reversely tracked along the dangerous defects until the leaf nodes, and the original key defects causing material failure are identified.
8. The system is characterized by comprising an image acquisition module, an image segmentation module, an image conversion module, a transformation module, an optimal candidate defect acquisition module and a failure defect determination module;
the image acquisition module is used for acquiring two-dimensional tomographic images of a plurality of transient states in the metal material sample;
the image segmentation module segments defects from the two-dimensional tomographic image based on a neural network and threshold segmentation, and generates a binary image only containing the defects and the background;
the image conversion module uses a three-dimensional space clustering method to represent defects in the binary image as three-dimensional point clouds;
the transformation module is used for identifying all possible target candidate defects of the tracked defects in a set range of adjacent states based on the defects represented by the three-dimensional point cloud through rotation and translation, carrying out the same rotation and translation on the environment information of the source object, obtaining a target candidate defect set of each tracked defect, and calculating an optimal rotation matrix and a translation vector corresponding to each candidate defect;
the optimal candidate defect acquisition module is used for determining an optimal candidate defect according to the similarity by calculating the similarity of the tracked defect and the target candidate defect environment point cloud;
and the failure defect determining module reversely establishes the defect composition of the final fracture of the material in each tensile state in a tree topology mode according to the target candidate defect set, the optimal rotation matrix and translation vector corresponding to each candidate defect and the optimal candidate defect, and reversely determines the original key defect which is finally failed.
9. A computer device comprising a processor and a memory, the memory storing a computer executable program, the processor reading the computer executable program from the memory and executing the computer executable program, the processor executing the computer executable program to implement the method for forward tracing and reverse fault localization of internal defects of metal according to any one of claims 1 to 7.
10. A computer readable storage medium, wherein a computer program is stored in the computer readable storage medium, and when the computer program is executed by a processor, the method for forward tracking and reverse fault localization of internal defects of a metal according to any one of claims 1 to 7 can be realized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311488555.0A CN117635543B (en) | 2023-11-09 | 2023-11-09 | Forward tracking and fault reverse positioning method and system for internal defects of metal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311488555.0A CN117635543B (en) | 2023-11-09 | 2023-11-09 | Forward tracking and fault reverse positioning method and system for internal defects of metal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117635543A true CN117635543A (en) | 2024-03-01 |
| CN117635543B CN117635543B (en) | 2024-04-30 |
Family
ID=90032970
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311488555.0A Active CN117635543B (en) | 2023-11-09 | 2023-11-09 | Forward tracking and fault reverse positioning method and system for internal defects of metal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117635543B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118037721A (en) * | 2024-04-11 | 2024-05-14 | 四川省嘉万光通信有限公司 | Method and system for detecting defects of optical cable harness |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170372099A1 (en) * | 2016-06-23 | 2017-12-28 | Carl Zeiss 3D Automation Gmbh | Carrier for the examination of workpieces by computed tomography |
| WO2020133639A1 (en) * | 2018-12-29 | 2020-07-02 | 东北大学 | Intelligent analysis system for magnetic flux leakage detection data in pipeline |
| US20220198647A1 (en) * | 2021-02-09 | 2022-06-23 | Nanjing University Of Aeronautics And Astronautics | Method for detecting and recognizing surface defects of automated fiber placement composite based on image converted from point cloud |
| CN115035081A (en) * | 2022-06-23 | 2022-09-09 | 西安交通大学 | Metal internal defect danger source positioning method and system based on industrial CT |
| CN116824294A (en) * | 2023-06-06 | 2023-09-29 | 中南大学 | Solder defect detection method and system based on small target identification |
-
2023
- 2023-11-09 CN CN202311488555.0A patent/CN117635543B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170372099A1 (en) * | 2016-06-23 | 2017-12-28 | Carl Zeiss 3D Automation Gmbh | Carrier for the examination of workpieces by computed tomography |
| WO2020133639A1 (en) * | 2018-12-29 | 2020-07-02 | 东北大学 | Intelligent analysis system for magnetic flux leakage detection data in pipeline |
| US20220198647A1 (en) * | 2021-02-09 | 2022-06-23 | Nanjing University Of Aeronautics And Astronautics | Method for detecting and recognizing surface defects of automated fiber placement composite based on image converted from point cloud |
| CN115035081A (en) * | 2022-06-23 | 2022-09-09 | 西安交通大学 | Metal internal defect danger source positioning method and system based on industrial CT |
| CN116824294A (en) * | 2023-06-06 | 2023-09-29 | 中南大学 | Solder defect detection method and system based on small target identification |
Non-Patent Citations (1)
| Title |
|---|
| 马永康;周涛;胡昌华;: "控制系统设计与分析中的潜通路分析技术", 电光与控制, no. 04, 30 August 2006 (2006-08-30), pages 97 - 100 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118037721A (en) * | 2024-04-11 | 2024-05-14 | 四川省嘉万光通信有限公司 | Method and system for detecting defects of optical cable harness |
| CN118037721B (en) * | 2024-04-11 | 2024-06-11 | 四川省嘉万光通信有限公司 | Method and system for detecting defects of optical cable harness |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117635543B (en) | 2024-04-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ali et al. | Structural crack detection using deep convolutional neural networks | |
| CN109523521B (en) | Pulmonary nodule classification and lesion positioning method and system based on multi-slice CT image | |
| Xu et al. | Stacked Sparse Autoencoder (SSAE) based framework for nuclei patch classification on breast cancer histopathology | |
| CN107154043B (en) | Pulmonary nodule false positive sample inhibition method based on 3DCNN | |
| Alsmadi et al. | Fish recognition based on robust features extraction from size and shape measurements using neural network | |
| Cao et al. | Collaborative discrepancy optimization for reliable image anomaly localization | |
| Wang et al. | Weakly supervised learning for whole slide lung cancer image classification | |
| Pan et al. | Mitosis detection techniques in H&E stained breast cancer pathological images: A comprehensive review | |
| Pan et al. | Cell detection in pathology and microscopy images with multi-scale fully convolutional neural networks | |
| Han et al. | Parts4Feature: Learning 3D global features from generally semantic parts in multiple views | |
| Lei et al. | Attention-guided multi-branch convolutional neural network for mitosis detection from histopathological images | |
| CN117635543B (en) | Forward tracking and fault reverse positioning method and system for internal defects of metal | |
| Karthiga et al. | Transfer learning based breast cancer classification using one-hot encoding technique | |
| Nateghi et al. | Automatic detection of mitosis cell in breast cancer histopathology images using genetic algorithm | |
| Das et al. | Contour-aware residual W-Net for nuclei segmentation | |
| Bogacz et al. | Period classification of 3D cuneiform tablets with geometric neural networks | |
| Karambakhsh et al. | SparseVoxNet: 3-D object recognition with sparsely aggregation of 3-D dense blocks | |
| Hao et al. | VP-Detector: A 3D multi-scale dense convolutional neural network for macromolecule localization and classification in cryo-electron tomograms | |
| Abed | Lung Cancer Detection from X-ray images by combined Backpropagation Neural Network and PCA | |
| Wu et al. | Fast particle picking for cryo-electron tomography using one-stage detection | |
| Xiong et al. | Attention u-net with feature fusion module for robust defect detection | |
| Gan et al. | Application study of PointNet++ and hybrid filtering based point cloud denoising and segmentation algorithms for Blisk | |
| Jing et al. | Defect segmentation with local embedding in industrial 3D point clouds based on transformer | |
| Boumaraf et al. | Conventional Machine Learning versus Deep Learning for Magnification Dependent Histopathological Breast Cancer Image Classification: A Comparative Study with Visual Explanation. Diagnostics, 2021; 11 (3): 528 | |
| Ma et al. | Visual detection of cells in brain tissue slice for patch clamp system |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |