Machine learning algorithm-based analysis of the crashworthiness of bio-inspired Al/PP sandwich tubes: Experimental and numerical investigation
Fahad M. Alhomayani, Husan Ali, Vahed Saif, Naser Mazraoui, Sameer Alghanmi, Hind Albalawi
Abstract
This research investigates the crashworthiness behavior of bio-inspired, hybrid aluminum and polypropylene (Al/PP) sandwich tubes, as well as single Al and PP hollow tubes, under quasi-static lateral loading both experimentally and numerically. It examines the effects of material permutation of single tubes and variations in inner tube diameters on energy absorption behavior by validating a Finite Element (FE) model against experimental results using LS-DYNA software for simulations. The simulation process accurately predicted the folding mechanism and crashworthiness parameters. The study then conducts a comprehensive FE parametric analysis, performing many simulations with variations in inner tube diameter and thickness, using a full-factorial design approach. To evaluate the tubes’ energy absorption, the research integrates the findings into machine learning algorithms. Specifically, Multilayer Perceptron models trained with the Bayesian-Regularization training algorithm are used to predict the specific energy absorption (SEA) and peak crushing force (PCF) to refine geometric parameters. The objective is to minimize PCF while maximizing SEA. Optimization using the non-dominated sorting genetic algorithm II (NSGA-II) identified an APA sandwich tube configuration with 1.5 mm thickness, 35 mm diameter and 8 core tubes as the optimal design for crash resistance.