2022 Schedule

Displaying 1 - 1 of 1

Dynamic P-F Curve with Machine Learning for efficient Predictive Maintenance

Presenter: Arun Gowtham
Reliability Engineer, RAMwright Consulting Co.
Organizations are observing the change in the maintenance landscape with the use of data-driven analytics for decision-making. Underpinning these analytics is Machine Learning. The algorithms form the model that ingests data to represent the system and predict its future state. While this method has found rapid applicability in other sectors, the field of Reliability & Maintenance Engineering is still exploring ways to adapt this idea to its conventional Asset Management programs. The objective of this paper is to explain the predictive power of machine learning by wrapping it around a prevalent reliability tool: the P-F Curve. Initially proposed in the Reliability-Centered Maintenance (RCM) framework, the P-F Curve is ubiquitous and the practitioners understand its simple and elegant description of the failure behavior. In spite of its understanding, the use of the P-F Curve has been minimal in everyday analysis to estimate when and how soon the failure will occur. Predictive Maintenance (PdM) Tools such as Vibration Analysis, Ultrasound, and Thermography have made the P-F Curve more accessible. Adding Machine Learning to these PdM tools, with a real-time data stream, will amplify the value of this analysis with better detection of Potential Failure (Pf) and forewarning of Functional Failure (Ff). Having real-time data and a real-time P-F Curve plot will enable the users to capture the changing conditions. We define this new curve as the Dynamic P-F Curve™. Dynamic P-F Curve™ Machine Learning models will estimate the time available (P- F Interval) for the maintenance team to respond to an asset before catastrophic failure. This interval will change if the asset experiences an external force causing it to wear out sooner. Thus, a dynamic curve makes the maintenance plan itself changeable and agile, improving the plan's efficiency. The final part of the presentation will showcase an example of how a Dynamic P-F Curve™ is calculated and represented by using an open-source data set. The resulting change in the maintenance plan actions is also prescribed to fully explain this idea, concluding with the list of use cases.
About the Presenter:

Arun Gowtham is a Certified Reliability Engineer with 7 years of industry experience applying Reliability Centered Maintenance (RCM) and Design for Reliability (DfR) practices to improve System Reliability. His breadth of exposure to diverse teams spanning 3 countries and 3 industries has given him unique insights & opportunities to tackle field failures and build sustaining reliability programs. His current interest is in employing data-driven analytics to underpin decision making for Reliability Management. Arun started reliability journey with a thesis work on Predictive Modeling for RCM as part of his master's degree graduation from Drexel University, PA, USA. He also holds a bachelor’s degree in Mechanical Engineering from Anna University, TN, India. Arun is an active volunteer; General Secretary for PEMAC GTA Chapter; and a registered engineer in Ontario.