Data-Driven Prognostics for Run-To-Failure Data Employing Machine Learning Models

The growth of prognostics study in manufacturing and engineering sectors has been rapidly increasing in recent years. With proper usage of IoT, big data and predictive analytics, all industries aim to improvise productivity, product lifetime and durability, ultimately improving customer satisfaction...

Full description

Saved in:
Bibliographic Details
Published in2020 International Conference on Inventive Computation Technologies (ICICT) pp. 528 - 533
Main Authors Saranya, E, Sivakumar, P. Bagavathi
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.02.2020
Subjects
Computational modeling
data-driven techniques
Handheld computers
Industrial Internet of Things (IIoT)
Industries
Machine learning
Machine learning (ML)
Manufacturing
Predictive Maintenance (PdM)
Predictive models
Productivity
Remaining Useful Lifetime (RUL)
Online AccessGet full text

Cover

More Information
Summary:The growth of prognostics study in manufacturing and engineering sectors has been rapidly increasing in recent years. With proper usage of IoT, big data and predictive analytics, all industries aim to improvise productivity, product lifetime and durability, ultimately improving customer satisfaction. Predicting Remaining Useful Lifetime (RUL) for deteriorating machines is helpful in avoiding unnecessary stoppages in production or usage of a product. This work focuses on understanding the scope and significance of prediction of RUL and use of machine learning models to achieve results of good precision. A comparative study on pros and cons of regression versus classification models is presented. In addition, the interpretation of the dataset used, influence of the features to the RUL estimation, significance of the presence of outliers in such run-to-failure data is examined.
DOI:10.1109/ICICT48043.2020.9112411