Failure strength prediction of glass/epoxy composite laminates from acoustic emission parameters using artificial neural network

Kumar, C. Suresh; Arumugam, V.; Sengottuvelusamy, R.; Srinivasan, S.; Dhakal, Hom Nath

doi:10.1016/j.apacoust.2016.08.013

Cited by 74 publications

(13 citation statements)

References 32 publications

(21 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In fact, these three features and their cumulative representations have been widely used to describe the failure process of the structure as well as the degree of damage accumulation. 29,32,35,40,51 Such a phenomenon is also originated from their definitions, that is, both are calculated by integrals. Another phenomenon consistent with their definitions is that AbE is slightly more sensitive to damage accumulation than E and SS.…”

Section: Discussionmentioning

confidence: 99%

“…Typical algorithms include classification and regression trees, 31,33 Multilayer Perception, 33 least-squares support vector machines, 34 support vector regression, 16 and the artificial neural network. 35 However, it is time-consuming and labor-intensive to construct a training library which contains a lot of labeled data. Moreover, whether the characteristics of the labeled signals from a certain AE source are consistent with those from the same source in actual composite structures remains unknown.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Dynamic feature evaluation on streaming acoustic emission data for adhesively bonded joints for composite wind turbine blade

Liu

Chen

et al. 2021

Structural Health Monitoring

View full text Add to dashboard Cite

Damage mode identification and premature failure prevention for composite structures by acoustic emission have drawn a great deal of attention. Feature evaluation on streaming acoustic emission data is one of the significant issues in research of acoustic emission signal processing. This work conducts dynamic feature evaluation on 15 conventional acoustic emission features so as to seek a deeper insight into different features with damage accumulation. First, the procedure of dynamic feature evaluation is presented based on three basic algorithms. Second, the streaming acoustic emission data are collected from the adhesively bonded composite single-lap joint subjected to quasi-static tensile loads. Third, further efforts are made so as to explore the information contained as well as to interpret the effect of damage accumulation. It is found that different conventional acoustic emission features show distinctive functions, including damage mode identification, damage process indication, and both of them. Informative features for damage pattern recognition are independent on damage accumulation. Useful features for damage process description show sensitive dynamic characteristics with damage accumulation, especially before the complete failure of the specimen. Furthermore, dynamic feature evaluation can be used to detect singular signals.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Dynamic feature evaluation on streaming acoustic emission data for adhesively bonded joints for composite wind turbine blade

Liu

Chen

et al. 2021

Structural Health Monitoring

View full text Add to dashboard Cite

show abstract

“…In this study, among various ML techniques, ANN was selected to build a prediction model from given experimental data sets for lap shear strength at room temperature and impact peel strength at −40 • C because ANN is the most effective technique for classifying and predicting complex nonlinear or linear relationships among numerous variables [20,21,23,24]. In the input data set for the ANN model, lap shear strength, and impact peel strength were set as target variables, and two separate ANN models were constructed to have only one target variable.…”

Section: Machine Learning Proceduresmentioning

confidence: 99%

Prediction of Lap Shear Strength and Impact Peel Strength of Epoxy Adhesive by Machine Learning Approach

et al. 2021

View full text Add to dashboard Cite

In this study, an artificial neural network (ANN), which is a machine learning (ML) method, is used to predict the adhesion strength of structural epoxy adhesives. The data sets were obtained by testing the lap shear strength at room temperature and the impact peel strength at −40 °C for specimens of various epoxy adhesive formulations. The linear correlation analysis showed that the content of the catalyst, flexibilizer, and the curing agent in the epoxy formulation exhibited the highest correlation with the lap shear strength. Using the analyzed data sets, we constructed an ANN model and optimized it with the selection set and training set divided from the data sets. The obtained root mean square error (RMSE) and R2 values confirmed that each model was a suitable predictive model. The change of the lap shear strength and impact peel strength was predicted according to the change in the content of components shown to have a high linear correlation with the lap shear strength and the impact peel strength. Consequently, the contents of the formulation components that resulted in the optimum adhesive strength of epoxy were obtained by our prediction model.

show abstract

“…The comparison between the supervised k-NN classifier and the unsupervised Kohonen’s map classifier provided a good agreement for discrimination of damage modes. 19–21 Yan et al 22 showed the significant potential of the self-organization neural network technique to classify AE signals. Self-organizing map (SOM) combined with k-means clustering was found to give the best clustering quality with low computational efforts for classifying peak frequency distribution of each damage mode.…”

Section: Introductionmentioning

confidence: 99%

Optimization of acoustic emission parameters to discriminate failure modes in glass–epoxy composite laminates using pattern recognition

Pabitha

Ramasamy

et al. 2018

Structural Health Monitoring

View full text Add to dashboard Cite

In this article, to overcome the challenges encountered during the discrimination of various failure modes in post impacted/indented glass-fiber-reinforced plastic, techniques like pattern recognition method and advanced signal processing were employed. The significant acoustic emission parameters such as amplitude, rise time, counts, energy, duration, and peak frequency that are acquired during compression after impact test are considered as inputs to cluster validity index and for various clustering techniques such as k-means, fuzzy C-means, and Kohonen's self-organizing map. The acoustic emission count-frequency and amplitude-frequency have no overlapping, whereas other combinations of acoustic emission parameters result in overlapping with four clusters. The clustering techniques are validated by discrete wavelet transform of acoustic emission signals. The discrete wavelet transform was performed on the clustered acoustic emission signals to identify the percentage of energy and frequency content of each level which correlates the different failure modes. The results infer that k-means, fuzzy C-means clustering, and Kohonen's self-organizing map are 94.5%, 97.1%, and 98.6% reliability, respectively, clearly suggesting Kohonen's self-organizing map as the most appropriate technique for the classification of acoustic emission signature.

show abstract

Failure strength prediction of glass/epoxy composite laminates from acoustic emission parameters using artificial neural network

Cited by 74 publications

References 32 publications

Dynamic feature evaluation on streaming acoustic emission data for adhesively bonded joints for composite wind turbine blade

Dynamic feature evaluation on streaming acoustic emission data for adhesively bonded joints for composite wind turbine blade

Prediction of Lap Shear Strength and Impact Peel Strength of Epoxy Adhesive by Machine Learning Approach

Optimization of acoustic emission parameters to discriminate failure modes in glass–epoxy composite laminates using pattern recognition

Contact Info

Product

Resources

About