Development of Key Performance Metrics for Aircraft Maintenance Process Control – CL10-LOCK
June 15, 2015
Productivity Improvements to Direct Mail Operations – LH08-TRAN
June 15, 2015

Syndromic Surveillance Systems – YR3 UA-AFRL 2075


Air Force Research Laboratory

Research Team:

Justin R. Chimka, James Burgmeier, Lauren Holloway, Heather L. Nachtmann, Patricia Cabrera, Manuel D. Rossetti

Universities Involved:

University of Arkansas

Start Date:


End Date:



This project addresses the issue of high profile disease outbreaks by better choice of data series and detection algorithms developing enhanced surveillance systems.
Prognostics involve prediction of the future state of a system. Typical prognostics systems involve sensors, data acquisition, and integrative software to synthesize and analyze data and report results with little or no human intervention. A good prognostic system provides an accurate picture of component degradation enabling appropriate PM actions to avoid catastrophic damage on critical parts and to maintain readiness rates for the system of interest. Much work has been performed in developing predictive algorithms to identify existing patterns between descriptive data of functioning systems and later malfunctioning systems and to predict malfunctioning events prior to occurrence. The human interface is missing from this research. Research questions of interest include: What is the role of the technician in this decision making process? How can the technician be empowered to improve prognostic predictions?
There is knowledge to be gleamed from other prognostic environments that can be adapted to improve the human interaction within equipment prognostics. These include physicians dealing with medical problems and weather forecasters projecting flood warnings. Physicians are charged daily with predicting patient prognoses based on problem detection and diagnosis. The human element of patient prognostics has been vastly studied. Researchers at the University of Virginia are developing models to help weather forecasters to make better flood predictions through the use of Bayesian, multi-objective, and utility theory techniques. These models focus on the forecaster’s decision process, so that the forecaster can “learn” to make better forecasts for flood warnings. Such systems enhance the decision-maker’s ability to make and maintain prognostic capabilities.
One way to explore the current human interaction within equipment prognostics is to examine historical prognostics decisions through technician interviews. This will provide first hand knowledge of the human side of equipment prognostics. Hypothetical equipment prognostics cases will then be developed and presented to MX technicians to observe their prognostic decision making process in a more controlled manner. The collected information will be qualitatively coded to provide insight into the current prognostic decision process that is being employed by MX technicians, i.e. what information do they use, what information is ignored, how often do they rely on past experience, confidence in their prognosis, etc.
Once this decision reasoning process is better understood, analysis on how to improve the equipment prognostics system from the viewpoint of the technician user will be performed. The system needs to provide recommended actions/alternatives and the implications of each recommended action to the user. Recommendations must be based on pertinent operational history, current and future mission objectives, and constraints. Guidance in the way this information should be presented to the technician will be sought through the examination of use cases designed to improve the information exchange.