Skripsi
ANALISA SISTEM KESEIMBANGAN VERTIKAL PADA WAHANA BAWAH AIR MENGGUNAKAN METODE ARTIFICIAL NEURAL NETWORK (ANN).
ABSTRACT The study analyzes the effectiveness of Artificial Neural Network (ANN) methods for vertical balance control in underwater vehicles (ROV). Traditional PID control methods were compared with ANN to evaluate performance differences. The research aims to identify the PID values required for training the ANN and to assess if ANN can achieve better balance control than PID in underwater systems. Data collection involved testing with sensors to gather a dataset comprising angular positions and control efforts. The ANN was trained using this dataset to predict and control the vehicle's balance more accurately.The results indicate that while PID control can stabilize the vehicle, it often results in oscillations around the zero angle before achieving balance. In contrast, ANN demonstrates superior performance by predicting and controlling motor actions more effectively, allowing the vehicle to reach and maintain the zero angle directly and accurately. The ANN's prediction and calculation capabilities enable smoother and more precise balance control compared to PID.However, implementing ANN on hardware like Arduino presents challenges due to memory limitations and slower processing speeds, which can hinder ANN's full potential. Despite these constraints, the study concludes that ANN offers a more accurate balancing process, with a standard deviation of 6.238 compared to PID's 7.545. The research highlights the potential of ANN in improving control systems for underwater vehicles, providing a foundation for further development in this field. Keywords: Artificial Neural Network (ANN), PID control, vertical balance, Remotely Operated Vehicle (ROV)
No other version available