REVIEW ON DATA INTEGRITY IN REMAINING USEFUL LIFE PREDICTION

Abstract

ABSTRACT - Maintenance has always been an important function in any manufacturing operation.
The main role of maintenance is to ensure that the manufacturing activities run at optimum condition
without any interruption. Several maintenance strategies have been introduced such as breakdown
maintenance, preventive maintenance, and condition-based maintenance. Condition-based
maintenance (CBM) has become one of the popular maintenance strategies in manufacturing
industries. CBM is a maintenance program that is triggered by the machine deterioration condition.
CBM serves two distinct purposes: diagnostic and prediction. The predictive segment largely focuses on
forecasting the remaining useful life (RUL) of components. Presently, the focus of study is primarily on
the prediction of Remaining Useful Life (RUL). A point often overlooked, the effectiveness of this
maintenance strategy depends on the quality and integrity of observation data. There have been a few
review papers on the broad scope of RUL prediction particularly in diagnostic techniques, but they lack
focus on the quality and integrity of the maintenance data. This paper provides a review on the
significance of data integrity in RUL. This review covers publications from 2005 to 2024. It offers a
categorisation of research fields or topics concerning data integrity and methods for data processing.
Deep learning neural network models especially LSTM and CNN have been used extensively to treat
data integrity issues, particularly in have also gained popularity, especially in newer publications. This
review provides evidence on the importance of data integrity in RUL prediction, and it can be useful for
new researchers in the area.