DESIGN OPTIMISATION OF METASTRUCTURE CONFIGURATION FOR ELECTRIC VEHICLE BATTERY PROTECTION USING MACHINE LEARNING METHOD

• Main Author: Cahyani Fatiha, Indira
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/73099
• Institution: Institut Teknologi Bandung

Electric vehicles are estimated to reach a shared market volume of $869 by 2027 due to their growing usage and popularity in recent years. However, this rapid growth has simultaneously emphasized the importance of battery safety during its application by developing a lightweight protection system with excellent energy absorption capabilities. Another development regarding the structure of electric vehicles is the effort to reduce weight. Through the design optimisation process, several meta structure cell configurations were assessed. Meta structure material is chosen for its negative Poisson ratio, which yields better energy absorption. In this research, three meta structure configurations (Bi-stable, Star-shaped, Double-U) have their geometry variables (thickness, inner spacing, cell stack) and material types (stainless steel, aluminum, and carbon steel) altered until the maximum Specific Energy Absorptions (SEA) value was attained. To simulate the mechanics of impact and determine the SEA of the varied designs, the Finite Element Method (FEM) is used. The design variation is generated using Latin Hypercube Sampling (LHS) to distribute the variables into 100 samples. The optimisation is then completed using the Machine Learning method to forecast the design that produces maximum SEA. Model approximation using Artificial Neural Networks (ANN) and variable optimisation using Nondominated Sorting Genetic Algorithm-II (NSGA-II) are the two steps that make up the machine learning optimisation process. The optimum control variables are Star-shaped cells consisting of one vertical unit cell using Aluminium material with the cross-section’s thickness =2.921 ???????? and the inner spacing of the cell =10.201 ????????. Comparing the optimum design to the baseline model, the SEA is increased by 5577%. The outcome is also validated using numerical simulation. The approximation model can accurately predict the most optimum design given the prediction error is only 2.83%. Four different sandwich structure configurations are then created using the optimum design. The optimised sandwich structure has the optimum design arranged in 6×4×1 cells resulting in a total dimension of 165.4×111.2×30 ???????? and the mass of 0.258 ????????. With a maximum deformation of 7.33 ????????, which is below the deformation threshold for prismatic battery failure, the optimised sandwich structure can prevent excessive deformation that causes battery failure from ground impact.