A general prognostic tracking algorithm for predictive maintenance

Prognostic Health Management (PHM) is a technology that uses objective measurements of condition and failure hazard to adaptively optimize a combination of availability, reliability, and total cost of ownership of a particular asset. Prognostic utility for the signature features are determined by transitional failure experiments. Such experiments provide evidence for the failure alert threshold and of the likely advance warning one can expect by tracking the feature(s) continuously. Kalman filters are used to track changes in features like vibration levels, mode frequencies, or other waveform signature features. This information is then functionally associated with load conditions using fuzzy logic and expert human knowledge of the physics and the underlying mechanical systems. Herein is the greatest challenge to engineering. However, it is straightforward to track the progress of relevant features over time using techniques such as Kalman filtering. Using the predicted states, one can then estimate the future failure hazard, probability of survival, and remaining useful life in an automated and objective methodology.