%A T. Meeks
%A A. Alqarni
%A S. Ponce
%A H. Fehr
%A H. Razip
%A O.P. Yadav
%A M.H. Hasan
%I Institute of Industrial and Systems Engineers, IISE
%T Evaluation of Machine Learning Algorithms in Predicting Corrosion Rates on Oil Refinery Piping
%D 2022
%J IISE Annual Conference and Expo 2022
%L scholars17551
%O cited By 0; Conference of IISE Annual Conference and Expo 2022 ; Conference Date: 21 May 2022 Through 24 May 2022; Conference Code:182057
%X Maintenance on piping failures due to degrading materials are time consuming, costly, and result in decreased productivity to fix. However, with growing technology it is possible to use machine learning models to predict when maintenance would be needed prior to such failures. With machine learning models and time series data analysis being a key aspect in the growing field of predictive maintenance, data from sensors in pipelines on corrosion were analyzed in Python to create a predictive model that would accurately predict plant equipment failure. Both classical and machine learning models were analyzed to use on the time series data of corrosion rates and Integrity Operating Window (IOW) tags collected from oil refinery piping. Models such as ARIMA and ARMA were comparatively the best at predicting the corrosion rates. On the other hand, machine learning models such as Random Forest (RF) and Support Vector Machine (SVM) were not as accurate, although RF was one of the more accurate machine learning models tested. Analyzing the corrosion rates yielded more accurate predictions than analyzing the IOW tags. Analyzing the IOW tags with machine learning models yielded percentage errors of 99 and above. However, ARIMA was found to have the highest success rate in predicting the corrosion rates obtained from the data. © 2022 IISE Annual Conference and Expo 2022. All rights reserved.
%K Condition based maintenance; Decision trees; Forecasting; Learning algorithms; Learning systems; Pipeline corrosion; Refining; Support vector machines; Time series; Time series analysis, Classical modeling; Corrosion prediction; Learning time; Machine learning algorithms; Machine learning models; Oil refineries; Operating windows; Random forests; Time series data analysis; Time series forecasting, Corrosion rate