Discovering and Explaining Patterns in Industrial Multivariate Time Series Data

Abstract: Complex assets and process units exhibit many different behaviors during the course of industrial operations. Identifying and removing sources of inefficiency in these operations is essential for advancing manufacturing and process operations. In this talk, we explain how classification as opposed to anomaly detection and forecasting is the essential machine learning problem for Industry 4.0. We explain the main challenges for these machine learning problems to motivate research directions. We then describe a signal processing pipeline and user interface for democratizing such machine learning and real-time processing.

Biography: Dr. Nikunj founded Falkonry after realizing that very valuable operational data produced in industrial infrastructure goes mostly unutilized in the energy, manufacturing and transportation sectors. Falkonry has enabled companies to scale predictive operations. Falkonry has significantly improved their uptime, yield and quality. Prior to Falkonry, Dr. Mehta led software architecture and customer success for C3 IoT. Earlier, he led innovation teams at Oracle focused on database technology and led the creation of the IndexedDB standard for databases embedded inside all modern browsers. He has contributed to standards at both W3C and IETF, and is a member of the ACM.

—

Host: Paul Bogdan

Sponsored by:

Center for Cyber-Physical Systems and the Internet of Things (CCI) http://cci.usc.edu
Ming Hsieh Institute for Electrical and Computer Engineering (MHI) http://mhi.usc.edu