Verification of Application of ANN Modelling in Study of Compressive Behaviour of Aluminium Sponges
Autor
Stręk, Anna Małgorzata
Dudzik, Marek
Kwiecień, Arkadiusz
Wańczyk, Krzysztof
Lipowska, Barbara
Opublikowane w
Engineering Transactions
Numeracja
Vol. 67, No 2
Strony
271–288
Data wydania
2019
Miejsce wydania
Warszawa
Wydawca
Polish Academy of Sciences Institute of Fundamental Technological Research, Université de Lorraine, Poznan University of Technology
Język
angielski
DOI
10.24423/EngTrans.991.20190615
Słowa kluczowe
metal sponges, aluminium sponges, compression tests, artificial neural networks
Abstrakt
This article presents a preliminary neural network analysis of the compressive behaviour of aluminium open-cell sponges and answers the question of whether this phenomenon can be modelled using artificial intelligence. The research consisted of two phases: first – compression experiments, which in turn provided data for the second phase – the artificial neural network (ANN) analysis. A two-argument function was proposed and tested using the gathered experimental data with a two-layer feedforward network. The determination coefficient R² for linear correlation between targets and modelling outputs was chosen as the criterion for the assessment of the quality of modelling. The obtained values were R²>0.96, which shows that neural networks hold the capacity to address the characterisation of the mechanical response of aluminium open-cell sponges in compression. Additionally, the mean absolute relative error (MARE) and the mean square error (MSE) were also determined.
