Creep modelling of polypropylenes using artificial neural networks trained with Bee algorithms

Esen, Ismail; D�?enci, Muharrem; D�?enci, Muharrem; Aydemir, Alpay; Esen, ?smail; Aydin, Mehmet Emin

Abstract

© 2015 Elsevier Ltd. All rights reserved. Polymeric materials, being capable of high mouldability, usability of long lifetime up to 50 years and availability at low cost properties compared to metallic materials, are in demand but finite element-based design engineers have limited means in terms of the limited material data and mathematical models. In particular, in the analysis of products with complex geometry, the stresses and strains of various amounts formed in the product should be known and evaluated in terms of a precise design of the product to fulfil life expectancy. Due to time and cost constraints, experimental data cannot be available for all cases required in analysis, therefore, finite element method-based simulations are commonly used by design engineers. This is also computationally expensive and requires a simpler and more precise way to complete the design more realistically. In this study, the whole creep behaviour of polypropylene for all stresses were obtained with 10% accuracy errors by artificial neural networks trained using existing experimental test results of the materials for a particular working range. The artificial neural network model was trained with traditional as well as heuristic based methods. It is demonstrated that heuristically trained ANN models have provided much accurate and precise results, which are in line with 10% accuracy of experimental data.

• Journal Article Type: Article
• Publication Date: Jan 1, 2015
• Deposit Date: Jun 23, 2015
• Publicly Available Date: Aug 10, 2016
• Journal: Engineering Applications of Artificial Intelligence
• Print ISSN: 0952-1976
• Publisher: Elsevier
• Peer Reviewed: Peer Reviewed
• Volume: 45
• Pages: 71-79
• DOI: https://doi.org/10.1016/j.engappai.2015.06.016
• Keywords: creep, polypropylene, artificial neural networks, bees algorithms, heuristically trained neural networks
• Public URL: https://uwe-repository.worktribe.com/output/841749
• Publisher URL: http://dx.doi.org/10.1016/j.engappai.2015.06.016
• Additional Information: Additional Information : © 2015, Elsevier. Licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International http://creativecommons.org/licenses/by-nc-nd/4.0/
• Contract Date: Aug 10, 2016