This project is one of major projects I accomplished at Utrecht University cooperated with The European Organization for Nuclear Research (CERN), Netherlands. The main goal is the devise a predictive system to warn the operators beforehand that the beam might decay, as well as to get better insight into what causes these dacays.
In this project, I am mainly focused on building and testing XGBoost model. My main jobs are:
• Improved Neural Network model performance by 20% of using Exponential Rolling Average method to remove stochastic noise from sensor data.
• Increased prediction accuracy by 40% of applying XGBoost (Extreme Gradient Boosting) to devise predictive system to warn operators beforehand beam current might decay.
• Captured change trend of beam current focusing on three sensitive regions and provided actual possible solutions for operators to highly reduce beam decay.