Selecting Classical Statistic or Bayesian/Markov based Diagnostic/Prognostics in Support Systems

Current and future needs must be met when implementing a diagnostic/ prognostic support system. The classical statistic commonly referred to as frequentists methodology, and the Bayesian methodology offer two different, but distinct, means to realize adaptive diagnostic/prognostic systems. The need for these systems is partially a consequence of the operational test program set (OTPS) design criteria not aligning with the field usage. The problem of selecting the classical statistic or Bayesian support system presents itself when applying the two methodologies in the support system paradigm. Modeling and simulating these two methodologies as they are applied within the support system paradigm provides an insight on their strengths and weaknesses. Dynamic test sequencing is examined as a means to improve the OTPS time to fault callout and improve the accuracy in identifying the repair action accuracy given the fault callout. Using Gaussian distributions [4] and density functions and beta distribution and density functions to model the process provides a means to quantify and leverage knowledge. Additionally, a Markov model is presented to demonstrate applicability to the support system paradigm. What exactly are these two methodologies, what are the differences and what are the pros and cons of each methodology. Systems with the ability to adapt and adjust to their environment are required to optimize OTPS performance by compensating for inaccurate fault callouts in line replaceable units (LRU) built in test as well as OTPS performance test, self-test, and diagnostic tests. The impacts to the way OTPSs are implemented and the impact to the legacy compilers are discussed. Virtual instrumentation and IVI and PnP drivers impacts are addressed and the direction which system architecture requirements need to migrate is addressed.