(UL)
  • Modelling creep behavior of PEEK at different temperatures using artificial neural networks [Elektronski vir]
    Aulova, Alexandra ; Bek, Marko
    Using neural networks (NN) for real-time structural health monitoring of polymer-based structures has advantages compared to existing methods. While existing structural health monitoring systems deal ... mostly with geometrical changes (cracks, delamination) NN can be trained to detect changes in material properties due to environmental conditions and time. With this contribution we demonstrate the capabilities of Multilayer Perceptron (MLP) neural network for response prediction of time- and temperature-dependent high performance polyether ether ketone (PEEK) at different temperatures. We investigated two different shear stress loading conditions at different temperatures: constant shear stress rate [1] and harmonical shear stress excitation. Training data was obtained experimentally by means of rotational rheometer. This contribution covers experimental procedure for training data generation and neural network design steps, including training procedure, choice of data for better generalization and topology optimization. We have shown that proper choice of temperatures corresponding to training data is of crucial importance for behavior prediction quality.
    Source: AERC 2022 – IBEREO 2022 [Elektronski vir] : 15th Annual European Rheology Conference [and] VIII Iberian Meeting on Rheology, Seville, Spain April 26-28, 2022 : book of abstracts (Str. 98)
    Type of material - conference contribution ; adult, serious
    Publish date - 2022
    Language - english
    COBISS.SI-ID - 114095875

    Link(s):

    https://aercsevilla2022.es/static/upload/ow130/events/ev408/AERC2022_Book_of_Abstracts_VF5.pdf

Shelf entry