OPTIMIZATION AND EXPERIMENTAL VALIDATION OF THIN WALL BEAM COMPOSITE UNDERGOING FOUR POINT BENDING CRUSH LOAD USING ARTIFICIAL NEURAL NETWORK (ANN) AND MACHINE LEARNING

• Main Author: Dimas Nur Hadian Shah, Wahyu
• Format: Theses
• Language: Indonesia
• Subjects: Teknik (Rekayasa, enjinering dan kegiatan berkaitan)
• Online Access: https://digilib.itb.ac.id/gdl/view/56944
• Institution: Institut Teknologi Bandung

This research discusses about subfloor structure optimization on conventional aircraft using Carbon Fiber Reinforced Plastic/Polymer (CFRP). CFRP is used because of its properties which is lighter but without compromising the strength of the structure significantly so that it can be used in various industries. CFRP usage as primary material in various industries especially in aerospace industry has increased the performance of structure compared with any conventional material such as metal. However, besides the material usage, optimization process can also be done by changing the configuration of composite material or optimize the topology of the structure to get the optimal performance. To get the configuration that can give the optimal performance of the structure we need to use machine learning (ML). Machine learning focuses on program that can learn from existing data so that it can be learned and be used as a basis to make a decision or to predict in which will be more accurate over time. Machine learning has been used in various industries, not only in engineering industries but also has been widely used in other industries such as banking, economics, etc. Machine learning itself has various methods including Artificial Neural Network (ANN) and Non-dominated Sorting Genetic Algorithm-II (NSGA-II). ANN is machine learning method that adopts the system of human nervous system which consists of several neurons where each neuron is connected to each other. Each neuron is associated with a weight that determines gow much influence that neuron has on the output of the system. NSGA-II is a machine learning concept inspired by the theory of natural evolution. In the theory of evolution, the weak species will be eliminated while the strong species will survive and pass their genes onto the next generation. Furthermore, species with the right combination of genes will become increasingly powerful and dominant in their population. In this research, we try to optimize how to achieve strength and bending strength as high as possible while also minimize the weight of the structure and also to optimize the energy absorption. Several parameters that will be varied are the cross-section of the structure, the orientation of composite fibres, the number of composite layers, and the size of cross-section of structure. By the end of this research, we found that the optimize configuration for the structure is I-beam cross section with the size of 44 mm, 16 layers of composite and orientation of composite fibre of 90o. It can be found that the configuration could improve the structure performance from the baseline model for each parameter are Strength per weight ratio up to 4.57%, Moment per weight ratio up to 5.36%, and Energy absorption up to 11.23%