PREDIKSI LAJU KOROSI ATMOSFERIK SENG DENGAN PEMBELAJARAN MESIN MENGGUNAKAN MODEL XGBOOST REGRESSOR

• Main Author: Nur Azizah, Jasmita
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/56482
• Institution: Institut Teknologi Bandung

Zinc coating on steel products is widely used in the automotive industry. Zinc was chosen for its corrosion resistance and sacrifical protection. In these applications zinc is a metal that is directly exposed to the atmospheric environment. Measurement of the atmospheric corrosion rate is carried out to determine the level of corrosion in zinc products exposed to the existing environment (brownfield) so that mitigation can be carried out to avoid initial failure of the product and in a new environment (greenfield) predictions can be made as an early stage or design stage in material selection. Exposure testing is a common test to study the behavior and value of the atmospheric corrosion rate for a long period of time and requires a long time. Therefore, the data based on the test is used as a sustainability analysis so that the analysis of corrosion data requires more advanced data mining methods. Machine learning was chosen as the problem method. In this study, the algorithm model used is Bagging Regressor, XGBoost Regressor, and Gradient Boosting. Hyperparameter tuning can optimize the algorithm model. The results obtained indicate that the XGBoost Regressor algorithm is the most optimal model based on the accuracy value and prediction validation results. This study also obtained predictions of atmospheric corrosion rates in Jakarta, Bandung, Valparaiso, Barcelona, and Japan successively the values are 10,619 g/m2; 9.589 g/m2; 7.813 g/m2; 2,943 g/m2; 11.132 g/m2.