Autonomous Learning for Condition Based Maintenance

Our STTR project will develop a data driven prognostic system that uses automated learning algorithms with stochastic artificial intelligence models to provide advanced warning of failure, fault, and other error events. Our work is based on new theory forimplementing learning algorithms within Bayesian stochastic models that have been developed by computer scientists at the University of New Mexico Artificial Intelligence Group. Bayesian learning is a key enabling technology for accurate autonomous realtime situation assessment from operating signatures of operating equipment. Management Sciences has teamed with UNM to develop and demonstrate a library of predictive engines based on self-learning used with advanced pattern recognition techniques toidentify the early signs of malfunctioning in operating machinery and electronic systems. The predictive engines will be commercialized in Phase II. Autonomous assessment through automated learning will provide breakthroughs for situation awarenessneeded for precise dynamic control, accurate condition assessment, self directed maintenance and precision logistics. The ability to predict machine/equipment events has significant commercial potential in aircraft, power, manufacturing, processing,transportation, and other industrial applications where such capability would allow companies to improve reliability and safety, reduce downtime, and lower the direct maintenance cost of physical assets.