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

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

    Research output: Contribution to journalArticlepeer-review

    364 Downloads (Pure)

    Abstract

    The ageing effect of glass/epoxy composite laminates exposed to seawater environment for different periods of time was investigated using acoustic emission (AE) monitoring. The mass gain ratio and flexural strength of glass fiber reinforced plastic (GFRP) composite laminates were examined after the seawater treatment. The flexural strength of the seawater treated GFRP specimens showed a decreasing trend with increasing exposure time. The degradation effects of seawater are studied based on the changes in AE signal parameters for various periods of time. The significant AE parameters like counts, energy, signal strength, absolute energy and hits were considered as training data input. The input data were taken from 40% to 70% of failure loads for developing the radial basis function neural network (RBFNN) and gener-
    alised regression neural network (GRNN) models. RBFNN model was able to predict the ultimate failure strength and could be validated with the experimental results with the percentage error well within 0.5–7.2% tolerance, whereas GRNN model was able to predict the ultimate failure strength with the percentage error well within 0.5–4.4% tolerance. The prediction accuracy of GRNN model is found to be better than RBFNN model.
    Original languageEnglish
    Pages (from-to)32-41
    Number of pages10
    JournalApplied Acoustics
    Volume115
    Early online date20 Aug 2016
    DOIs
    Publication statusPublished - 1 Jan 2017

    Keywords

    • GFRP composite laminates
    • acoustic emission
    • seawater degredation
    • artificial neural network
    • RBFNN
    • GRNN

    Fingerprint

    Dive into the research topics of 'Failure strength prediction of glass/epoxy composite laminates from acoustic emission parameters using artificial neural network'. Together they form a unique fingerprint.

    Cite this