Logistics Community Meeting

When:  Apr 20, 2021 from 12:00 to 13:00 (ET)
Associated with  MORS Community

The Logistics Community of Practice will meet on 20 April at 1200 ET! Our speaker will be Dr. Sanja Cvijic, presenting "Probabilistic Programming Approaches to Predictive Maintenance and Information-based Logistics (PMIL)."

 

Maintaining mission readiness while also reducing lifecycle costs remains an ongoing challenge and top priority across the DoD. Efficient energy resource planning and maintenance scheduling require predictive data-driven decision-support tools that can inform decision makers on what actions are needed and when. In this talk, two related problems in terms of logistics planning and the probabilistic programming tools will be introduced.

 

The first problem is a Predictive Condition Based Maintenance problem in two distinct domains: cables on board operationally active vessels and power transformers in distribution power grids. Although the domains are vastly different, predictive maintenance challenges and solutions are very similar. For cable health monitoring and fault detection, we will introduce a Distributed Analysis Tool for Enterprise Monitoring – Cable Calibration Tool (DATEM-CCT) tool which detects, classifies, and localizes faults in Ship Signals Exploitation Equipment (SSEE) Inc-F Cable Calibration test results. For power transformer health monitoring, fault prediction and estimation of Remaining Useful Lifetime (RUL), we will introduce Probabilistic Operations Warranted for Energy Reliability Evaluation and Diagnostics (POWERED) tool which uses electrical, thermal and Dissolved Gas Analysis (DGA) data to monitor, detect and classify power transformer faults.

 

The second logistics planning problem is efficient planning of Operational Energy (OE) related to the effective management of energy and supply resources including energy and fuel for power systems, generators and weapon platforms. Complex real-world systems and their energy consumption needs are very hard to model given complex relationships among types of missions, weather and the headcount. We will demonstrate Energy Models of Critical Components (E-MC2) tool which uses probabilistic programming to construct and learn models of complex real-worlds for resource-monitoring and alerting of potential resource consumption issues.

 

Dr. Sanja Cvijic is a probabilistic research scientist at Charles River Analytics. She is involved in developing probabilistic models and applying machine learning techniques for data-driven decision support across various domains. She is the PI on Probabilistic Operations Warranted for Energy Reliability Evaluation and Diagnostics (POWERED) project, leading the design of a health and monitoring tool for power transformers. Prior to joining Charles River Analytics in 2020, Dr. Cvijic’s work was focused on the modeling, control and optimization of power networks. Her research includes the use of Big Data techniques state estimation and adaptive Alternating Current Optimal Power Flow (AC OPF) optimization, which were deployed to the New York Power Pool (NYPA). She has also designed frameworks for modeling and tracing loop flows, which were then applied for flow control. She received PhD in Power Systems in Electrical and Computer Engineering from Carnegie Mellon University in 2013 and BS in Computer Science from University of Belgrade, Serbia in 2008. She has co-authored many journal and conference publications and has two patents.



Link: https://www.gotomeet.me/MORSMeeting50a/logistics
Phone: +1 (872) 240-3212
Access Code: 577-956-821

Contact

Elizabeth Marriott
(703) 933-9073
liz.marriott@mors.org