VALIDASI DAN OPTIMISASI KOLOM HIBRIDA ALUMINUM-KOMPOSIT BERDINDING TIPIS DENGAN SUDUT JAMAK PADA KASUS PEMBEBANAN AKSIAL MENGGUNAKAN METODE MACHINE LEARNING

• Main Author: Fikri Ramdani, Mohamad
• Format: Theses
• Language: Indonesia
• Subjects: Teknik (Rekayasa, enjinering dan kegiatan berkaitan)
• Online Access: https://digilib.itb.ac.id/gdl/view/57050
• Institution: Institut Teknologi Bandung

Technological advances have led to the emergence of various new materials. Energy efficiency requirements as well as reducing carbon emissions are driving factors for creating lightweight and efficient materials. One of them is Fiber Metal Laminates (FML). FML are hybrid composite structures consisting thin sheets of metal alloys and layers of fiber-reinforced polymer materials. FML are made to combine the advantages of metals and composites. Based on several studies that have been conducted, composite metal hybrid materials have been shown to increase crashworthiness performance compared to metals and lower costs than composites. Machine Learning is rapidly developing to be used to identify complex data structures that are usually nonlinear. With the use of machine learning, an accurate prediction model can be generated. In this study, validation and optimization of the crashworthiness performance of thin-walled aluminum composite hybrid column with multicorner was carried out in the case of axial loading with an approximation function approached by artificial neural network (ANN), then the optimization process was carried out using the nondominated sorting genetic algorithm II (NSGA-II). The validation process shows a good correlation between the numerical simulation and the experiment results. For the aluminum model the average error is 4%. Meanwhile in the hybrid model, the average error for orientation [30/-30]s, [45/-45]s, and [60/-60]s were 9%, 6%, and 8%, respectively. Furthermore, the numerical simulation model can describe the physical phenomena that occur well. From the verification process that has been carried out, it is found that the optimization model made can predict well with the largest error value in the SEA parameter of 12%. Meanwhile, when compared to the Baseline Model, the optimal model can provide a significant increase in the Specific Energy Absorption (SEA), Crushing Force Efficiency (CFE), dan Mean Crushing Force (MCF) parameters by 5%, 36%, and 37%. Meanwhile, the decrease in the Peak Force parameter is only 0.4%.