This grant provides funding for the development of theory and algorithms necessary to establish effective prognostics methods and technologies for Condition-Based Maintenance (CBM) of critical equipment and systems through statistical and computational intelligence methods and sensor-fusion technology. The objective is to develop prognostic algorithms that facilitate accurate estimation of Remaining-Useful-Life (RUL) values and confidence levels for components and equipment. The prognostic methods for signal modeling and forecasting will exploit existing theory in three promising areas: Statistical Nonlinear Mixed-Effect Models, Dynamic Bayesian Network Models, and Structural-Learning Neural Networks. In order to accurately establish a definition of failure in the degradation signal domain, several probabilistic and fuzzy-inference methods will also be developed to accommodate physics based failure mechanisms and experiential and/or empirical knowledge. The resulting methods and technology will be further refined and validated using a Computer Numerical Control (CNC) machining test-rig.
If successful, the results of this research will lead to a better understanding for the science of equipment prognostics (often regarded as the Achilles' heel of CBM) and a methodology to work with it in a generic framework. The generic nature of these algorithms could lead to a more widespread permeation of CBM technology, thus bringing down the cost of maintenance operations while avoiding unnecessary maintenance. Another benefit is improved availability of equipment (resulting from reduced or eliminated equipment breakdowns). The resulting methods and technology are relevant for almost all industries.
