Maintenance of critical or/complex systems has recently moved from traditional preventive maintenance to Condition Based Maintenance (CBM) exploiting the advances both in hardware (sensors/DAQ cards, etc.) and in software (sophisticated algorithms blending together the state of the art in signal processing and pattern analysis). Along this path, Environmental Control Systems and other critical systems/processes can be improved based on concepts of anomaly detection, fault diagnosis and failure prognosis.

This paper presents an empirical model to describe battery behavior during individual discharge cycles as well as over its cycle life. The basis for the form of the model has been linked to the internal processes of the battery and validated using experimental data. Subsequently, the model has been used in a Particle Filtering framework to make predictions of remaining useful life for individual discharge cycles as well as for cycle life. The prediction performance was found to be satisfactory as measured by performance metrics customized for prognostics.

Opportunities exist to apply nonlinear filtering to model-based prognostics in order to provide a systematic way of dealing with the propagation of system damage at some future time, whenever imprecise diagnostic information is obtained. Central to the prognostics problem is the ability to properly capture and manage uncertainties when predicting remaining useful life of a particular component of interest. The goal of this paper is to present a foundation for prediction and filtering of the failure process using nonlinear prognostic models and exact (finite-dimensional) filters.

Model-based prognostics exploits domain knowledge of the system, its components, and how they fail by casting the underlying physical phenomena in a physics-based model that is derived from first principles. In most applications, uncertainties from a number of sources cause the predictions to be inaccurate and imprecise even with accurate models. Therefore, algorithms are employed that help in managing these uncertainties. Particle filters have become a popular choice to solve this problem due to their wide applicability and ease of implementation.

Particle filters (PF) have been established as the de facto state of the art in failure prognosis, and particularly in the representation and management of uncertainty in long-term predictions when used in combination with outer feedback correction loops. This paper presents a novel Risk-Sensitive PF (RSPF) framework that complements the benefits of the classic approach, by representing the probability of rare and costly events within the formulation of the nonlinear dynamic equation that describes the evolution of the fault condition in time.