CN114925716A - Carbon fiber composite material damage positioning method based on integrated learning algorithm - Google Patents
Carbon fiber composite material damage positioning method based on integrated learning algorithm Download PDFInfo
- Publication number
- CN114925716A CN114925716A CN202210287705.0A CN202210287705A CN114925716A CN 114925716 A CN114925716 A CN 114925716A CN 202210287705 A CN202210287705 A CN 202210287705A CN 114925716 A CN114925716 A CN 114925716A
- Authority
- CN
- China
- Prior art keywords
- learning algorithm
- model
- carbon fiber
- damage
- fiber composite
- 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
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B7/00—Measuring arrangements characterised by the use of electric or magnetic techniques
- G01B7/16—Measuring arrangements characterised by the use of electric or magnetic techniques for measuring the deformation in a solid, e.g. by resistance strain gauge
- G01B7/18—Measuring arrangements characterised by the use of electric or magnetic techniques for measuring the deformation in a solid, e.g. by resistance strain gauge using change in resistance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Medical Informatics (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Investigating Strength Of Materials By Application Of Mechanical Stress (AREA)
Abstract
The invention discloses a carbon fiber composite material damage positioning method based on an integrated learning algorithm, which comprises the following steps of (1) selecting a tested area of a CFRP (carbon fiber reinforced polymer) experiment and constructing a strain sensor test network; (2) sequentially applying mass blocks with different weights to the CFRP to simulate damage, and acquiring strain characteristic signals; (3) constructing an integrated learning algorithm model, and training the integrated learning algorithm model; (4) and realizing damage positioning by using the obtained optimal ensemble learning model. The invention integrates and combines a plurality of weak learners to form a strong learner, thereby effectively improving the prediction performance of the learner.
Description
Technical Field
The invention relates to the technical field of damage positioning of carbon fiber composites, in particular to a carbon fiber composite damage positioning method based on an integrated learning algorithm.
Background
Carbon fiber Composite (CFRP) materials are widely used in the fields of rail transit, aerospace, and the like because of their advantages of light weight, high strength, good fatigue resistance, and the like. In recent years, CFRP has been drawing attention from countries around the world as a new alternative material. If the CFRP is used as a key stressed part and bears a large external load for a long time, the internal fiber bundle is broken, and the damage of the CFRP in the running process of a railway vehicle or a space plane inevitably causes huge casualties and property loss, so people hope to utilize sensor technology to combine measurement data to evaluate the health state of the current CFRP.
At present, a great deal of research is carried out on damage localization and pattern recognition of materials, wherein the research carried out by a machine learning algorithm comprises the following steps: zhang Yanjun and the like measure the static strain of 304 steel plates in a damage mode by using a strain sensor, and are applied to identification of steel plate damage positions by combining a particle swarm algorithm and a least square support vector machine algorithm. A composite material impact load real-time online monitoring system is constructed by utilizing the Lujiyun and the like, and a carbon fiber composite material structure impact positioning method based on wavelet packet feature extraction and a support vector regression machine is researched. And constructing a Fiber Bragg Grating (FBG) sensor network by using the Soxhlet and the like, and simultaneously exploring the damage mode identification problem of the CFRP by combining a wavelet decomposition and reconstruction algorithm, spectral analysis and a support vector multi-classifier. The method for identifying and monitoring the damage of the steel structure is constructed by combining the fully-connected neural network model and the transfer rate function. JANG and the like train a neural network to establish a nonlinear relation between an impact response signal and the position of an impact point by collecting a large number of impact point response signal samples, and input the response signal of the impact point to be detected into a training model to realize the positioning of the impact point. In summary, the composite material damage positioning and identifying research mainly realizes prediction according to a single model, and the prediction precision of the single model cannot be guaranteed.
Disclosure of Invention
The invention aims to provide a carbon fiber composite material damage positioning method based on an integrated learning algorithm.
In order to realize the purpose, the invention adopts the following technical scheme:
the carbon fiber composite material damage positioning method based on the integrated learning algorithm comprises the following steps:
(1) selecting a tested area of the carbon fiber composite material, establishing a coordinate system xoy in the tested area, equally dividing the tested area according to the coordinate system xoy to form a plurality of grid coordinates, and respectively pasting strain gauges in the directions of the x axis and the y axis of the coordinate system;
(2) selecting three weight blocks with known weights and different masses, taking any one weight block, sequentially placing the weight block in each grid coordinate to obtain a strain characteristic signal generated by the structural stress change of the carbon fiber composite material, wherein the three weight blocks can obtain three strain characteristic signal data sets and correspondingly record the three strain characteristic signal data sets as damage modes 1, 2 and 3;
(3) constructing an ensemble learning algorithm model, wherein the ensemble learning algorithm model comprises an SVR-Adaboost model taking a support vector regression as a weak learner, a random forest model taking a CART tree as the weak learner, an extreme random tree model and a gradient lifting decision tree model;
(4) inputting strain characteristic signals acquired in a damage mode 1 and grid coordinate data corresponding to the strain characteristic signals into the ensemble learning algorithm model constructed in the step (3) to train the ensemble learning algorithm model, wherein 70% of data are subjected to ensemble learning algorithm model training, and 30% of data are subjected to ensemble learning algorithm model testing; and inputting the strain characteristic signals acquired by the damage modes 2 and 3 into the trained ensemble learning algorithm model to realize damage positioning, and verifying that the ensemble learning model has transfer learning capability in the damage positioning application field.
Further, performing data normalization processing on the strain characteristic signals acquired in the step (2), wherein the data normalization processing comprises baseline drift elimination and noise reduction processing.
Further, the integrated learning algorithm model after training adopts the Pearson correlation coefficient and the root mean square error to predict the model accuracy.
Further, the number of weak learner iterations of the optimal model of the ensemble learning algorithm is 35.
The method can realize damage positioning by integrating a learning algorithm in the damage positioning of the carbon fiber composite material, and has the transfer learning capability.
Drawings
FIG. 1 is a schematic diagram of coordinate division of a region under test according to the present invention.
FIG. 2 is a diagram illustrating the relationship between the correlation coefficient and the iteration number of the weak learner.
FIG. 3 is a diagram illustrating the relationship between RMSE and weak learner iterations.
FIG. 4 is a schematic diagram of the relationship between the predicted global coordinate test error and the iteration number of the weak learner.
Fig. 5 is a schematic diagram illustrating comparison between the predicted coordinates and the actual coordinates according to the present invention.
Detailed Description
The method for positioning the damage of the carbon fiber composite material based on the integrated learning algorithm comprises the following steps:
(1) selecting a tested area of the carbon fiber composite material, establishing a coordinate system xoy in the tested area, equally dividing the tested area according to the coordinate system xoy to form a plurality of grid coordinates, and respectively pasting strain gauges in the directions of the x axis and the y axis of the coordinate system;
(2) selecting three weight blocks with known weights and different masses, taking one weight block and sequentially placing the weight block in each grid coordinate to obtain a strain characteristic signal generated by the structural stress change of the carbon fiber composite material, wherein the three weight blocks can obtain three strain characteristic signal data sets and correspondingly record the three strain characteristic signal data sets as damage modes 1, 2 and 3;
(3) constructing an ensemble learning algorithm model, wherein the ensemble learning algorithm model comprises an SVR _ Adaboost model taking a support vector regression machine as a weak learner, a random forest model taking a CART tree as the weak learner, an extreme random tree model and a gradient lifting decision tree model;
(4) inputting strain characteristic signals acquired in the damage mode 1 and grid coordinate data corresponding to the strain characteristic signals into the ensemble learning algorithm model constructed in the step (3) to train the ensemble learning algorithm model, wherein 70% of data are subjected to ensemble learning algorithm model training, and 30% of data are subjected to ensemble learning algorithm model testing; and inputting the strain characteristic signals acquired by the damage modes 2 and 3 into the trained ensemble learning algorithm model to realize damage positioning, and verifying that the ensemble learning model has transfer learning capability in the damage positioning application field.
The effectiveness of this example was verified experimentally.
To simulate the damage of CFRP under static force, the CFRP specification of 450mm × 450mm × 3mm was selected, and a 120mm × 120mm experimental area was selected, as shown schematically in FIG. 1. The test area for which the damage recognition is performed is indicated by a dashed line frame, and 2 uniaxial resistance strain gauges 1 and 2 are attached to the test area. The measured area was equally divided into 13 × 13 grid coordinates each having a length of 10mm, and there were 169 coordinate points after the division was completed. And (3) supporting the CFRP at four ends, sequentially applying masses of 400g, 500g and 600g to 169 coordinate points, and measuring the strain capacity of the CFRP to obtain 507 groups of strain characteristic data under three damage modes. The experimental working conditions of 400g, 500g and 600g are sequentially recorded as damage modes 1, 2 and 3, the same weight is applied to different positions to simulate different damage positions, and different weights are applied to the same position to simulate different damage modes. The change of CFRP structure stress caused by heavy loading influences the mechanical characteristics, and is a common damage simulation mode in simulation analysis and experimental tests. 169 groups of data are acquired in a single experiment, strain data are acquired by IMC data acquisition, and strain values are displayed on a computer in real time.
Due to the temperature rise effect of the strain gauge, data drift occurs in the measuring process, and the drift amount becomes larger along with the increase of time. In the experimental process, mutation points appear in a partial region of a measured signal, certain fluctuation exists in the signal after the signal is stable, certain interference exists in the selection of strain characteristic quantities, and baseline drift needs to be eliminated and the signal needs to be subjected to noise reduction processing in order to extract effective strain characteristics.
In order to realize the final coordinate prediction, the predicted x coordinate and the predicted y coordinate are respectively modeled by adopting the same model, two input and single output models are constructed, a support vector regression machine and a CART tree are sequentially established to train data, 135 groups of data are taken as a training set to carry out model training after 169 groups of data in the damage mode 1 are subjected to normalization treatment, and the rest 34 groups of data are taken as a test set to carry out model testing.
The kernel function of the support vector regression machine selects a Gaussian radial basis kernel function, a Python programming language is combined to realize the selection of an optimal relaxation variable gamma and a penalty strength C by a grid search method, and when the fitting effect of an x coordinate model is optimal (gamma is optimal) x ,C x ) When the y-coordinate model fit is optimal (γ) when (1.02,4.21) is true y ,C y )=(3.05,11.58)。
The CART tree takes the maximum tree depth as an optimization variable, the number of samples of the minimum segmentation node is set to be 2, the number of samples of the leaf node is set to be 1, and a square error loss function is selected to adjust the internal parameters of the regression tree. And when the fitting effect of the x and y coordinate models is optimal, the maximum tree depth is 10.
Training data by taking an optimal support vector regression model (SVR) as a weak learner of Adaboost, and simultaneously comparing the prediction precision of a Random Forest (RF), an Extreme Random Tree (ERT) and a Gradient Boosting Decision Tree (GBDT) which are formed by taking an optimal CART tree as the weak learner.
Fig. 2 shows the variation relationship between the correlation coefficient between the predicted x coordinate and the actual y coordinate of the four ensemble learning models and the iteration number of the weak learner on the test set. The data show that the correlation of x and y coordinates predicted by the SVR _ Adaboost model increases with the increase of the iteration number, the maximum relation number of the x axis is 0.9876, the maximum relation number of the y axis is 0.9848, the correlation coefficient tends to be stable after 10 iterations, the correlation coefficient of the Random Forest (RF) model for predicting the y axis coordinate tends to be stable after 22 iterations, the maximum correlation coefficient of the x axis coordinate is 0.9857, and the maximum correlation coefficient of the x axis coordinate is 0.9905. The coordinate correlation coefficient of an x-axis and a y-axis of a gradient lifting decision tree (GBDT) model is not obviously changed along with the increase of the iteration number, the maximum relation number of the x-axis is 0.9889, and the y-axis is 0.9711. The Extreme Random Tree (ERT) model predicts that the x-coordinate correlation coefficient is 0.9918 at most, the y-axis is 0.9853, and the correlation coefficient gradually increases with the increase of the iteration number and tends to be stable after 20 iterations. From the overall change trend, the ensemble learning model has the capability of improving the correlation between the predicted coordinates and the actual coordinates.
Fig. 3 shows the variation relationship between RMSE and weak learner iterations between the predicted values and the measured values of the test set. In combination with data, root mean square errors of x and y coordinates predicted by all models are gradually reduced and tend to be stable with the gradual increase of the iteration number, the RMSE is the minimum SVR _ Adaboost model, the x coordinate error is the minimum 6.64, and the y axis is the minimum 6.48. The root mean square errors of the predicted results of the three tree models of RF, GBDT and ERT are large, and the minimum root mean square errors of x and y axes of the three tree models are 10.04 and 11.61, and all the minimum root mean square errors appear in the random forest model. The general trend shows that the increase of the iteration number can reduce the prediction error of the model.
The abscissa and the ordinate are combined, a change relation between a test error and a weak learner iteration number when the four kinds of integrated learning models predict an integral coordinate on the same test set is given in fig. 4, data show that when the model iteration number reaches 35, prediction errors of all models gradually decrease and tend to a stable state, prediction errors of an Extreme Random Tree (ERT) and a Random Forest (RF) model are close to each other, prediction errors of an SVR _ Adaboost model and a Gradient Boosting Decision Tree (GBDT) model are close to each other, the prediction error of the RF model is 2.66%, the prediction error of the ERT model is 2.56%, the prediction error of the SVR _ Adaboost model is 3.45%, and the lowest prediction error of the GBDT model is 3.63%. The optimal model is an ERT model with 35 iterations. Fig. 5(a) shows the coordinate relationship between 10 predicted points and actual points on the test set after 35 iterations of the ensemble learning model weak learner.
And normalizing the data of the damage mode 2 and the damage mode 3, and randomly selecting 34 groups to be brought into the ensemble learning model of the weak learner for 35 iterations. In the prediction error of the damage mode 2, the prediction error of the GBDT model is the minimum and is 12.4%; the RF model predicted the largest error, 16.3%. Among the prediction errors of the damage pattern 3, the GBDT model prediction error is the smallest, 10.9%. The prediction error of the RF model is the largest and is 18.6%, and the integrated learning model still has higher prediction precision. Fig. 5(b) and 5(c) show coordinate diagrams of 10 predicted points in the damage modes 2 and 3.
In the method, the damage positioning can be realized by integrating the learning algorithm in the damage positioning of the carbon fiber composite material, and the method has the transfer learning capability.
The above description is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any modification and replacement based on the technical solution and inventive concept provided by the present invention should be covered within the scope of the present invention.
Claims (4)
1. The carbon fiber composite material damage positioning method based on the integrated learning algorithm is characterized by comprising the following steps of:
(1) selecting a tested area of the carbon fiber composite material, establishing a coordinate system xoy in the tested area, equally dividing the tested area according to the coordinate system xoy to form a plurality of grid coordinates, and respectively pasting strain gauges in the directions of the x axis and the y axis of the coordinate system;
(2) selecting three weight blocks with known weights and different masses, taking one weight block and sequentially placing the weight block in each grid coordinate to obtain a strain characteristic signal generated by the structural stress change of the carbon fiber composite material, wherein the three weight blocks can obtain three strain characteristic signal data sets and correspondingly record the three strain characteristic signal data sets as damage modes 1, 2 and 3;
(3) constructing an ensemble learning algorithm model, wherein the ensemble learning algorithm model comprises an SVR _ Adaboost model taking a support vector regression machine as a weak learner, a random forest model taking a CART tree as the weak learner, an extreme random tree model and a gradient lifting decision tree model;
(4) inputting strain characteristic signals acquired in the damage mode 1 and grid coordinate data corresponding to the strain characteristic signals into the ensemble learning algorithm model constructed in the step (3) to train the ensemble learning algorithm model, wherein 70% of data are subjected to ensemble learning algorithm model training, and 30% of data are subjected to ensemble learning algorithm model testing; and inputting the strain characteristic signals acquired by the damage modes 2 and 3 into the trained ensemble learning algorithm model to realize damage positioning, and verifying that the ensemble learning model has transfer learning capability in the damage positioning application field.
2. The integrated learning algorithm-based carbon fiber composite damage positioning method according to claim 1, characterized in that: and (3) carrying out data normalization processing on the strain characteristic signals acquired in the step (2), wherein the data normalization processing comprises baseline drift elimination and noise reduction processing.
3. The integrated learning algorithm-based carbon fiber composite damage positioning method according to claim 1, characterized in that: and the trained ensemble learning algorithm model adopts the Pearson correlation coefficient and the root mean square error to predict the model precision.
4. The integrated learning algorithm-based carbon fiber composite damage positioning method according to claim 3, characterized in that: the weak learner iteration times of the optimal model of the ensemble learning algorithm are 35.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210287705.0A CN114925716B (en) | 2022-03-22 | 2022-03-22 | Carbon fiber composite material damage positioning method based on ensemble learning algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210287705.0A CN114925716B (en) | 2022-03-22 | 2022-03-22 | Carbon fiber composite material damage positioning method based on ensemble learning algorithm |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114925716A true CN114925716A (en) | 2022-08-19 |
CN114925716B CN114925716B (en) | 2023-08-25 |
Family
ID=82805582
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210287705.0A Active CN114925716B (en) | 2022-03-22 | 2022-03-22 | Carbon fiber composite material damage positioning method based on ensemble learning algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114925716B (en) |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101794396A (en) * | 2010-03-25 | 2010-08-04 | 西安电子科技大学 | System and method for recognizing remote sensing image target based on migration network learning |
CN102495134A (en) * | 2011-12-08 | 2012-06-13 | 大连理工大学 | Truss structure damage positioning method based on added mass |
CN108520535A (en) * | 2018-03-26 | 2018-09-11 | 天津大学 | Object classification method based on depth recovery information |
CN110261487A (en) * | 2019-01-22 | 2019-09-20 | 华东理工大学 | A kind of damage detection apparatus System and method for of composite material pressure container |
CN111024821A (en) * | 2019-12-30 | 2020-04-17 | 大连理工大学 | Composite material storage box health monitoring system and method |
CN111582350A (en) * | 2020-04-30 | 2020-08-25 | 上海电力大学 | Filtering factor optimization AdaBoost method and system based on distance weighted LSSVM |
CN111652252A (en) * | 2020-06-11 | 2020-09-11 | 中国空气动力研究与发展中心超高速空气动力研究所 | Ultrahigh-speed impact damage quantitative identification method based on ensemble learning |
CN112989969A (en) * | 2021-02-26 | 2021-06-18 | 清远市智慧农业农村研究院 | Crop pest and disease identification method and device |
CN113554067A (en) * | 2021-07-01 | 2021-10-26 | 清华大学深圳国际研究生院 | Optical fiber winding defect detection method and device based on ensemble learning |
KR20220011456A (en) * | 2020-07-21 | 2022-01-28 | 울산과학기술원 | Method of detecting the location of damage for composite using machine learning |
CN114058691A (en) * | 2021-11-09 | 2022-02-18 | 山西医科大学 | Skeletal muscle early injury time prediction method based on Stacking ensemble learning |
-
2022
- 2022-03-22 CN CN202210287705.0A patent/CN114925716B/en active Active
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101794396A (en) * | 2010-03-25 | 2010-08-04 | 西安电子科技大学 | System and method for recognizing remote sensing image target based on migration network learning |
CN102495134A (en) * | 2011-12-08 | 2012-06-13 | 大连理工大学 | Truss structure damage positioning method based on added mass |
CN108520535A (en) * | 2018-03-26 | 2018-09-11 | 天津大学 | Object classification method based on depth recovery information |
CN110261487A (en) * | 2019-01-22 | 2019-09-20 | 华东理工大学 | A kind of damage detection apparatus System and method for of composite material pressure container |
CN111024821A (en) * | 2019-12-30 | 2020-04-17 | 大连理工大学 | Composite material storage box health monitoring system and method |
CN111582350A (en) * | 2020-04-30 | 2020-08-25 | 上海电力大学 | Filtering factor optimization AdaBoost method and system based on distance weighted LSSVM |
CN111652252A (en) * | 2020-06-11 | 2020-09-11 | 中国空气动力研究与发展中心超高速空气动力研究所 | Ultrahigh-speed impact damage quantitative identification method based on ensemble learning |
KR20220011456A (en) * | 2020-07-21 | 2022-01-28 | 울산과학기술원 | Method of detecting the location of damage for composite using machine learning |
CN112989969A (en) * | 2021-02-26 | 2021-06-18 | 清远市智慧农业农村研究院 | Crop pest and disease identification method and device |
CN113554067A (en) * | 2021-07-01 | 2021-10-26 | 清华大学深圳国际研究生院 | Optical fiber winding defect detection method and device based on ensemble learning |
CN114058691A (en) * | 2021-11-09 | 2022-02-18 | 山西医科大学 | Skeletal muscle early injury time prediction method based on Stacking ensemble learning |
Non-Patent Citations (5)
Title |
---|
HUAN LIU 等: "Data-Driven Approaches for Characterization of Delamination Damage in Composite Materials", IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, vol. 68, no. 3, pages 2532, XP011821844, DOI: 10.1109/TIE.2020.2973877 * |
刘宇韬 等: "基于AdaBoost-LSSVM的纤维复合材料损伤识别", vol. 46, no. 09, pages 148 - 153 * |
杨宇 等: "一种集成多个机器学习模型的复合材料结构损伤识别方法", vol. 35, no. 02, pages 278 - 287 * |
杨斌 等: "基于超声导波的压力容器健康监测Ⅲ:纤维缠绕压力容器的在线监测", 机械工程学报, vol. 56, no. 10, pages 19 - 26 * |
谢坤明: "基于随机森林算法的结构损伤识别", 福建交通科技, no. 05, pages 91 - 93 * |
Also Published As
Publication number | Publication date |
---|---|
CN114925716B (en) | 2023-08-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Jung et al. | Advanced deep learning model-based impact characterization method for composite laminates | |
US20130289953A1 (en) | Self-optimizing, inverse analysis method for parameter identification of nonlinear material constitutive models | |
CN107169205B (en) | Iron ore classification modeling method | |
Ozmen et al. | A novel methodology with testing and simulation for the durability of leaf springs based on measured load collectives | |
Bakir et al. | Correlation of simulation, test bench and rough road testing in terms of strength and fatigue life of a leaf spring | |
Chin et al. | Durability prediction of coil spring through multibody-dynamics-based strain generation | |
CN114065373A (en) | Automobile control arm rack endurance test method, device and equipment | |
Slimani et al. | Improved ANN for damage identification in laminated composite plate | |
Tong et al. | Research on random load characteristic based on mixture distribution model | |
CN114925716A (en) | Carbon fiber composite material damage positioning method based on integrated learning algorithm | |
Shin | Real-time recovery of impact force based on finite element analysis | |
CN113973403B (en) | Temperature-induced strain field redistribution intelligent sensing method based on structure discrete measurement point topology | |
CN114254533B (en) | Method for examining influence and prediction of fatigue vibration on fixed angle of product group component | |
CN112816122B (en) | Bolt tightness degree monitoring method based on deep learning and piezoelectric active sensing | |
Vettori et al. | A virtual sensing approach to operational modal analysis of wind turbine blades | |
CN113722308A (en) | Acceleration response data completion method and device based on EEMD-MultiCNN-LSTM | |
Kepka et al. | Practical notes for assessing the fatigue life of bodyworks of buses and trolleybuses | |
CN110487579A (en) | A kind of girder construction damnification recognition method based on inclination angle slope | |
Kelmar | Machine Learning Based Sensor Selection for Modal Testing | |
CN113688771B (en) | LNG storage tank acceleration response data supplementing method and device | |
CN113962134B (en) | Strain monitoring method for linear superposition of condition generation type countermeasure network and load strain | |
Candelieri et al. | A hyper-solution framework for SVM classification: Improving damage detection on helicopter fuselage panels | |
CN114925573B (en) | Subway sleeper beam damage data set construction method based on equivalent strain hypothesis | |
CN116663434B (en) | Whole vehicle load decomposition method based on LSTM deep neural network | |
Aguir et al. | Identification of constitutive parameters using hybrid ANN multi-objective optimization procedure |
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 |