CN113706511A - Composite material damage detection method based on deep learning - Google Patents
Composite material damage detection method based on deep learning Download PDFInfo
- Publication number
- CN113706511A CN113706511A CN202111013211.5A CN202111013211A CN113706511A CN 113706511 A CN113706511 A CN 113706511A CN 202111013211 A CN202111013211 A CN 202111013211A CN 113706511 A CN113706511 A CN 113706511A
- Authority
- CN
- China
- Prior art keywords
- damage
- composite material
- alexnet network
- deep learning
- detecting
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- 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
- 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
- G06N3/047—Probabilistic or stochastic 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
- G06N3/08—Learning methods
-
- 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
- G06T2207/10004—Still image; Photographic image
-
- 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
- G06T2207/20076—Probabilistic image processing
-
- 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
- G06T2207/20081—Training; Learning
-
- 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
Abstract
The invention provides a method for detecting damage of a composite material based on deep learning, which comprises the following steps of S1, pre-training an AlexNet network by using an image of an ImageNet database; s2, collecting a data set of the image of the composite material from the existing composite material literature; s3, randomly selecting 75% of images in the data set for AlexNet network training, and randomly selecting 25% of images for verifying the AlexNet network; and S4, adjusting the learning rate of the trained AlexNet network, and classifying the type and the severity of the damage of the composite material. According to the invention, the AlexNet network training model is adopted to replace a manual visual inspection and nondestructive inspection mode to detect the damage degree and the damage type of the composite material, so that the detection precision and efficiency are improved, and the detection cost is reduced.
Description
Technical Field
The invention relates to a method for detecting damage of a composite material based on deep learning, and belongs to the technical field of automatic visual detection.
Background
The composite material has the advantages of high specific strength, strong vibration damping capacity, wear resistance, fatigue resistance, high temperature resistance and the like. Due to these excellent properties, composite materials are widely used in various fields such as civil use, aerospace, automobiles, and the like. One disadvantage of composite materials is that they are susceptible to damage, which can lead to reduced fatigue life or catastrophic failure of the material. The ability to identify damage and measure the extent of damage by appropriate techniques is crucial to the integrity and performance assessment of composite materials, and it is therefore an urgent need to develop effective detection methods for damage to composite materials.
Most of the technical companies are also the damage detection of the composite material by manual visual inspection and nondestructive inspection techniques for technical reasons, and the detection methods have the following problems: (1) the manual visual inspection efficiency is low, the accuracy and the precision of detection are different from person to person, and workers are easily influenced by the external environment during detection, so that certain risks exist. (2) Nondestructive testing generally requires a relatively high cost and time, requires components to be stopped during inspection, and cannot distinguish the type and degree of damage.
Aiming at the defects, the invention provides the composite material damage detection method based on deep learning, which can replace manual visual inspection and nondestructive inspection, improve the detection precision and efficiency and reduce the detection cost.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a method for detecting damage of a composite material based on deep learning, which has the specific technical scheme that:
a method for detecting damage of a composite material based on deep learning comprises the following steps:
s1, pre-training an AlexNet network by using an image of an ImageNet database;
s2, collecting a data set of the image of the composite material from the existing composite material literature;
s3, randomly selecting 75% of images in the data set for AlexNet network training, and randomly selecting 25% of images for verifying the AlexNet network;
and S4, adjusting the learning rate of the trained AlexNet network, and classifying the type and the severity of the damage of the composite material.
Preferably, the AlexNet network architecture includes 5 convolutional layers, 3 pooling layers, 3 fully-connected layers, and 1 Softmax layer.
Preferably, the images collected from the composite literature in S2 are required to have different thickness, material, laminate and texture features.
Preferably, the types of images collected in S2 include undamaged, impact damaged, eroded, matrix cracked, and fiber broken.
Further, the types of impact damage include four types of low energy impact fronts, high energy impact fronts, low energy impact backs, and high energy impact backs.
Preferably, in S3, before training and verifying the AlexNet network, the image size is adjusted by using the enhanced image data storage algorithm to satisfy the condition of the AlexNet input layer, that is, the image size is 227 × 227 × 3.
Preferably, the initial learning rate for training the AlexNet network in S3 is adjusted to 0.0002.
Further, in S4, the learning rate of the trained AlexNet network is adjusted to 0.00005, so as to realize the damage classification of matrix cracking and fiber breakage; the learning rate of the trained AlexNet network is adjusted to 0.0001, so that impact damage, erosion and undamaged classification can be realized.
According to the invention, the AlexNet network training model is adopted to replace a manual visual inspection and nondestructive inspection mode to detect the damage degree and the damage type of the composite material, so that the detection precision and efficiency are improved, and the detection cost is reduced.
Drawings
FIG. 1 is a schematic flow chart of a method for detecting damage to a composite material based on deep learning according to the present invention.
Fig. 2 is an architecture diagram of an AlexNet network of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, a method for detecting damage to a composite material based on deep learning includes the following steps:
s1, pre-training an AlexNet network by using an image of an ImageNet database; the architecture of the AlexNet network includes 5 convolutional layers, 3 pooling layers, 3 fully-connected layers, and 1 Softmax layer, as shown in fig. 2.
S2, collecting a data set of images of the composite material from existing literature of laminated composite materials having different thicknesses, materials, laminates, textures, etc., the types of images including undamaged, impact damaged, eroded, cracked matrix and broken fibers, the types of impact damage being further divided into four types, a low energy impact front, a high energy impact front, a low energy impact back and a high energy impact back. Low energy impact and high energy impact damage types are visually distinguishable, with images of fiber breaks that are clearly visible being classified as high energy images. Low energy impact and high energy impact damage are specifically used to define the severity of damage to a composite material, low energy being expressed as weak severity and high energy being expressed as heavy severity.
The collected images comprise datasets of 20, 24, 16, 52, 25, 39, 28 and 24 images corresponding to matrix cracking, fiber breakage, undamaged, low energy impact front, high energy impact front, low energy impact back, high energy impact back, and erosion, respectively.
S3, randomly selecting 75% of images in the data set for AlexNet network training, and randomly selecting 25% of images for verifying the AlexNet network; first, the image size is adjusted using the enhanced image data storage algorithm to satisfy the condition of the AlexNet input layer (i.e., the image size is 227 × 227 × 3). Then, the AlexNet network was trained using these images with an initial learning rate of 0.0002. After training is completed, the performance of the network is evaluated by verifying AlexNet network classification.
And S4, adjusting the learning rate of the trained AlexNet network, and classifying the type and the severity of the damage of the composite material. Adjusting the learning rate of the trained AlexNet network to be 0.00005 so as to realize damage classification of matrix cracking and fiber breakage; the learning rate of the trained AlexNet network is adjusted to be 0.0001, so that the classification of impact damage, erosion and undamaged can be realized, and the severity of impact damage and the front and back of impact damage can be distinguished.
The invention adopts the AlexNet network model to detect and classify the damage types and the damage severity of the composite materials, replaces manual visual inspection and nondestructive inspection, improves the detection precision and efficiency and reduces the detection cost.
Although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that various changes in the embodiments and/or modifications of the invention can be made, and equivalents and modifications of some features of the invention can be made without departing from the spirit and scope of the invention.
Claims (8)
1. A method for detecting damage of a composite material based on deep learning is characterized by comprising the following steps: the method comprises the following steps:
s1, pre-training an AlexNet network by using an image of an ImageNet database;
s2, collecting a data set of the image of the composite material from the existing composite material literature;
s3, randomly selecting 75% of images in the data set for AlexNet network training, and randomly selecting 25% of images for verifying the AlexNet network;
and S4, adjusting the learning rate of the trained AlexNet network, and classifying the type and the severity of the damage of the composite material.
2. The method for detecting the damage of the composite material based on the deep learning of claim 1, wherein: the AlexNet network architecture includes 5 convolutional layers, 3 pooling layers, 3 fully-connected layers, and 1 Softmax layer.
3. The method for detecting the damage of the composite material based on the deep learning of claim 1, wherein: the images collected from the composite literature in said S2 are required to have different thickness, material, lamination and texture features.
4. The method for detecting the damage of the composite material based on the deep learning of claim 1, wherein: types of images collected in S2 include undamaged, impact damaged, eroded, matrix cracked, and fiber broken.
5. The method for detecting the damage of the composite material based on the deep learning as claimed in claim 4, wherein: the types of impact damage include four types, a low energy impact front, a high energy impact front, a low energy impact back, and a high energy impact back.
6. The method for detecting the damage of the composite material based on the deep learning of claim 1, wherein: in S3, before training and verifying the AlexNet network, the image size is adjusted by using the enhanced image data storage algorithm to satisfy the condition of the AlexNet input layer, that is, the image size is 227 × 227 × 3.
7. The method for detecting the damage of the composite material based on the deep learning of claim 1, wherein: the initial learning rate for training the AlexNet network in S3 is adjusted to 0.0002.
8. The method for detecting the damage of the composite material based on the deep learning of claim 7, wherein: in the step S4, the learning rate of the trained AlexNet network is adjusted to be 0.00005, so that damage classification of matrix cracking and fiber breakage is realized; the learning rate of the trained AlexNet network is adjusted to 0.0001, so that impact damage, erosion and undamaged classification can be realized.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111013211.5A CN113706511A (en) | 2021-08-31 | 2021-08-31 | Composite material damage detection method based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111013211.5A CN113706511A (en) | 2021-08-31 | 2021-08-31 | Composite material damage detection method based on deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113706511A true CN113706511A (en) | 2021-11-26 |
Family
ID=78658075
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111013211.5A Pending CN113706511A (en) | 2021-08-31 | 2021-08-31 | Composite material damage detection method based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113706511A (en) |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107292333A (en) * | 2017-06-05 | 2017-10-24 | 浙江工业大学 | A kind of rapid image categorization method based on deep learning |
CN108549852A (en) * | 2018-03-28 | 2018-09-18 | 中山大学 | Pedestrian detector's Auto-learning Method under special scenes based on the enhancing of depth network |
CN108614997A (en) * | 2018-04-04 | 2018-10-02 | 南京信息工程大学 | A kind of remote sensing images recognition methods based on improvement AlexNet |
CN108765391A (en) * | 2018-05-19 | 2018-11-06 | 科立视材料科技有限公司 | A kind of plate glass foreign matter image analysis methods based on deep learning |
CN109376636A (en) * | 2018-10-15 | 2019-02-22 | 电子科技大学 | Eye ground image classification method based on capsule network |
CN109508650A (en) * | 2018-10-23 | 2019-03-22 | 浙江农林大学 | A kind of wood recognition method based on transfer learning |
CN109522965A (en) * | 2018-11-27 | 2019-03-26 | 天津工业大学 | A kind of smog image classification method of the binary channels convolutional neural networks based on transfer learning |
CN111275677A (en) * | 2020-01-17 | 2020-06-12 | 哈尔滨工业大学 | Ceiling earthquake damage identification method based on convolutional neural network |
CN111462068A (en) * | 2020-03-30 | 2020-07-28 | 华南理工大学 | Bolt and nut detection method based on transfer learning |
CN111860596A (en) * | 2020-06-17 | 2020-10-30 | 长安大学 | Unsupervised pavement crack classification method based on deep learning and model establishment method |
CN113096088A (en) * | 2021-04-07 | 2021-07-09 | 浙江大学 | Concrete structure detection method based on deep learning |
-
2021
- 2021-08-31 CN CN202111013211.5A patent/CN113706511A/en active Pending
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107292333A (en) * | 2017-06-05 | 2017-10-24 | 浙江工业大学 | A kind of rapid image categorization method based on deep learning |
CN108549852A (en) * | 2018-03-28 | 2018-09-18 | 中山大学 | Pedestrian detector's Auto-learning Method under special scenes based on the enhancing of depth network |
CN108614997A (en) * | 2018-04-04 | 2018-10-02 | 南京信息工程大学 | A kind of remote sensing images recognition methods based on improvement AlexNet |
CN108765391A (en) * | 2018-05-19 | 2018-11-06 | 科立视材料科技有限公司 | A kind of plate glass foreign matter image analysis methods based on deep learning |
CN109376636A (en) * | 2018-10-15 | 2019-02-22 | 电子科技大学 | Eye ground image classification method based on capsule network |
CN109508650A (en) * | 2018-10-23 | 2019-03-22 | 浙江农林大学 | A kind of wood recognition method based on transfer learning |
CN109522965A (en) * | 2018-11-27 | 2019-03-26 | 天津工业大学 | A kind of smog image classification method of the binary channels convolutional neural networks based on transfer learning |
CN111275677A (en) * | 2020-01-17 | 2020-06-12 | 哈尔滨工业大学 | Ceiling earthquake damage identification method based on convolutional neural network |
CN111462068A (en) * | 2020-03-30 | 2020-07-28 | 华南理工大学 | Bolt and nut detection method based on transfer learning |
CN111860596A (en) * | 2020-06-17 | 2020-10-30 | 长安大学 | Unsupervised pavement crack classification method based on deep learning and model establishment method |
CN113096088A (en) * | 2021-04-07 | 2021-07-09 | 浙江大学 | Concrete structure detection method based on deep learning |
Non-Patent Citations (1)
Title |
---|
VICAVO: ""深度学习卷积后图像尺度的计算"", 《HTTP://BLOG.CSDN.NET.NET/VICAVO/ARTICLE/DETAILS/98305639》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Fotouhi et al. | Autonomous damage recognition in visual inspection of laminated composite structures using deep learning | |
Saeedifar et al. | Delamination analysis in composite laminates by means of acoustic emission and bi-linear/tri-linear cohesive zone modeling | |
CN107908879B (en) | Method for evaluating fatigue performance of concrete beam bridge | |
CN112268799A (en) | Static strength and fatigue strength integrated test verification method for composite material structure | |
CN101853320A (en) | Fuzzy comprehensive evaluation method suitable for aircraft structure corrosion damage | |
CN110879254A (en) | Steel rail crack acoustic emission signal detection method based on improved least square generation type countermeasure network | |
Sahoo et al. | Effect of patch-parent stacking sequence and patch stiffness on the tensile behaviour of the patch repaired carbon-glass hybrid composite | |
CN105004367B (en) | Monolithic integrated optical circuit storage life characteristic detection method | |
CN112199796A (en) | Method for designing composite material culvert casing layering | |
Ai et al. | Crack detection and evolution law for rock mass under SHPB impact tests | |
Cairns et al. | Wind turbine composite blade manufacturing: the need for understanding defect origins, prevalence, implications and reliability. | |
CN113706511A (en) | Composite material damage detection method based on deep learning | |
Nair et al. | Studying failure modes of GFRP laminate coupons using AE pattern-recognition method | |
CN109858152B (en) | Method for predicting residual fatigue life of composite material containing initial delamination damage | |
CN101236136B (en) | Curve plotting method characterizing aeroplane service lifetime degree and applications | |
CN110335242A (en) | A kind of tire X-ray defect detection method based on multi-model fusion | |
Smith et al. | Applications of ultrasonic NDT to aerospace composites | |
CN113051763A (en) | Ship system-level electronic equipment reliability evaluation method based on data fusion technology | |
CN112906263B (en) | Method for predicting strength of composite material laminated plate containing hole-making delamination damage | |
Yu et al. | A deep-learning-based method for damage identification of composite laminates [J] | |
Gift | A critical review of the role of acoustic emission algorithms in the delamination analysis of adhesive joints | |
Fotouhi et al. | Achieving robust acoustic emission-based damage characterisation of scaled laminated composites under indentation | |
Smith et al. | Modelling the mechanical properties of as-manufactured composite components based on 3D non-destructive characterisation | |
Dabetwar | Fatigue damage diagnostics and prognostics of composites using artificial intelligence methods for experimental data | |
Zhang et al. | Vibration monitoring of composite laminates with delamination damage |
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 |