Consulting Professor, Information Systems Laboratory
I advise graduate students, teach graduate courses, and lead projects.
My work is in analytics for Industrial Internet of Things (IoT) - mathematical algorithms that help making better decisions for data generated by machines.
A seminar class on Industrial Internet of Things is taught in Spring 2017; here is the Spring 2016 class. Seminar on Intelligent Energy Systems: Big Data was taught in 2015, 2014, 2013, 2012, and 2011. Past classes include Fault Diagnostics Systems in Spring 2009 as well as Control Engineering in Industry in Spring 2005 and in Winter 2003.
My current interest is in Industrial IoT (IIoT) analytics.
Most applicatons of Data Science and Machine Learning are to people-generated data, e.g., advertising.
Applications to machine-generated data, e.g., jet engine or electrial power system data are emerging.
This area is variously known as Industrial Internet, Internet of Things (IoT), Internet of Everything, and IIoT in the US industry. Closely reated are NSF Cyber-physical Systems research area and European Industry 4.0 initiative.
Analytical processing for mission-critical industrial applications is implemented in Operational Technology (OT) systems. The IIoT drives convergence of the OT and IT (Information Technology) systems. The IT analytics are known as operations research.
My earlier work was in OT analytics domains of decision & control and signal processing. The emerging IIoT analytics combine Machine Learning with methods of decision & control and operations research. This is a fruitful area for research and valuable industrial applications.
If you are a graduate student, who is interested in Industrial Internet of Things, please feel free to contact me. There might be an opportunity to work in this area.
My company Mitek Analytics LLC works with US Air Force, NASA, electrical utilities, airlines, and Fortune 500 companies. Previously, I spent a decade with Honeywell working on aircraft and space systems, turbomachines (jet engines), and process control applications. Earlier work includes automotive and robotics (force control of robotics systems, legged locomotion) applications.
Condition and Failure Analytics work includes collaboration with Professor Stephen Boyd. A NASA project looked at optimization methods for machine learning that scale engineering models for airline big data collected for fleets of assets such as aircraft and jet engines. A related Air Force project builds fleet reliability models from maintenance data. Earlier work included condition monitoring and fault isolation for Air Force jet engines in collaboration with GE Aviation as well as NSF project on distributed sensor monitoring in collaboration with Honeywell.
Data-driven Risk Analytics work includes collaboration with Professor Steven Chu. Big data sets can include enough extreme events to afford detailed statistical modeling based on the extreme value theory (EVT). One application is to trending the risk of extreme weather events in the changing climate using high-resolution geo-spatial historical weather. As one example, the risk of 100-year extreme hot weather events in the continental US was found to increase over 200% in the last 4 decades. Another application is to data analytics for computing resource demand in the cloud in collaboration with Yahoo. The peak demand events can result in violations of service level agreement (SLA). Data-driven optimization allowed achieving 70% improvement in the data center performance.
Smart Grid Analytics project was a collaboration with Professor Sanjay Lall. It was sponsored by Precourt Institute for Energy, where I am an Affiliated Faculty Member, and analyzed electrical distribution systems. Other related work includes probabilistic forecasting of operational grid loads and analyzing transmission planning requirements for regional power systems operator.