Perpustakaan Digital ITB

In 2019, air transportation played a critical role in the global economy, contributing over 35% to the total value of goods transported worldwide and accounting for 3.5% of the world's GDP. As a major contributor to the global economy, the air transportation sector faces substantial operational costs, with fuel consumption accounting for approximately 20% of these expenses. Therefore, effective fuel management is very crucial, as even minor deviations can significantly impact costs and operational reliability. Responding to the significance of fuel consumption in the aviation industry, this study investigates the application of a Multi-Layer Perceptron (MLP) neural network model to predict fuel consumption throughout various flight phases of a Boeing 747-300 aircraft. It alsp identifies key parameters influencing fuel consumption and evaluates the model's transferability across various routes using homogeneous data, underscoring its potential utility in diverse operational scenarios. By utilizing a comprehensive dataset encompassing 11 input parameters, with two hidden layers (10 and 5 neurons, ReLU activation) trained using the Adam optimizer and L2 regularization, the model achieved an R² of 0.998 on training and testing datasets. Validation on both homogeneous and heterogeneous datasets consistently yielded R² values above 0.995 across all flight phases. The study identifies engine RPM, rate of altitude change, and true airspeed as the most influential parameters in predicting fuel consumption, with engine RPM being the most critical predictor. These findings provide airlines with a robust, data-driven tool for precise fuel consumption prediction across diverse operational conditions, facilitating optimized fuel management and cost-saving strategies.