ATMOSPHERIC CORROSION RATE PREDICTION OF CARBON STEEL IN SOUTHEAST ASIA USING MACHINE LEARNING

• Main Author: Sanika Aulia, Irza
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/56479
• Institution: Institut Teknologi Bandung

Carbon steel, which is commonly used as a construction material, cannot be separated from atmospheric corrosion damage, especially in Southeast Asia. That region has a tropical climate with high temperature, relative humidity, wind speed, and other weather factors. Many researches are conducted to predict the corrosion rate so the problem can be prevented, but with conventional experciment method it will spend much time and cost. One of the methods to predict the corrosion rate is machine learning. To make an optimal model, first the environment factor and corrosion rate data were collected. Then the dataset was analyzed so the correlation between parameters was known. This machine learning process uses Gradient Boosting and Decision Tree Regression model, and their accuracy and error calculation are pretty similar. But from the validation model and prediction, Gradient Boosting is more sensitive with input data. The prediction of corrosion rate in several places in Southeast Asia are carried out, and the result of corrosion rate for Gresik 18,4; Bandung 15,7; Phuket 17,8; Phra pradaeng 16,0; and Yangon 16,1 all in ?m/year. To make the machine learning prediction more accurate, it is necessary to collect more good quality data.