Fault Probabilistic Inference Methodology

Fault Probabilistic Inference Methodology

Docket: UAH-P-16022


Currently, Condition Based Maintenance (CBM) of military helicopters are tracked by Condition Indicators (CI) calculated from special vibration sensors. These sensors are part of a system designed to reactively predict fault appearance. However, they are somewhat susceptible to corrupt data due to artifact and noise sensitivities.

Fault growth is an irreversible process comprised of rare events. Because the fault growth process is random in nature, in order to predict fault appearance, a numerical method must be equipped to handle sequential vibration data in the context of a fault evolution model. This way, the artifacts that inevitably distort the Power Spectral Density (PSD) can be effectively rejected.

Researchers at UAH have developed a fault probabilistic inference methodology for the prediction of fault, or crack, appearance. This method uses machine learning and analysis of probability data to determine the fault growth in a substance. In testing, this method obtained a data accuracy rate of over 95% over a 6-year time frame of tested military aircraft.


  • Defense
  • Materials testing
  • Reliability testing
  • Infrastructure


  • Significant reduction in defense fleet maintenance costs
  • Reduces NEOFs (false-positives)
  • High accuracy rate of over 95%


  • State of Development: Developed
  • Licensing Status: Available for licensing
  • Patent Status: Proprietary