Event Prognosis, Asset Reliability and Maintenance
The research direction aims at developing advanced methodologies for event data using IoT sensor data. An "event" marks a random elicitation in the time domain. In the context of machines, systems, and production equipment -- an event can be a warning, a failure, a maintenance activity, etc. Thus, one can easily find a system (or a fleet of systems) to experience a multitude of events during its operational lifetime. Some of the broad research goals under this direction are:
To predict the occurrence of a critical event in the presence of system heterogeneity, complex relationships, and unit censoring. We develop methods to share information between similar (yet different) units, learn the associations, and offer a real-time prognosis.
To model the interrelated effect of a variety of maintenance events on a component's lifetime. We develop methods to infer the impact of shared maintenance actions across components and use them to create dynamic and cost-efficient maintenance plans.
Related articles:[1]
To utilize sensor monitoring signals as well as discrete data from critical components (such as, batteries, etc.) and offers dynamic maintenance plans.
Domain-aware Active Learning
We create methods to synergistically combine domain knowledge, (i.e., biology and physics) with real-world experiments to develop next-gen bioprinted liver tissue models.
In collaboration with: Dr. Srikanthan Ramesh, Dr. Sundar Madihally.