Kurtosis as a Metric in the Assessment of Gear Damage

Gear diagnostics is becoming an important area of research, especially in critical applications such as rotorcraft propulsion where safety is paramount. Many rotorcraft statistics illustrate the need for dedicated monitoring systems to reliably diagnose the faults. A gear system fails when it ceases to efficiently perform the function for which it was designed. This can be the result of a single catastrophic event or an accumulation of initially undetected and rather innocuous events in the gear system. It is well known that evaluation of statistical properties will give reasonable diagnostic indication of gear damage. Although there are a large number of such statistical attributes such as root mean square value, crest factor, skewness, kurtosis, and so on, kurtosis has emerged as a single number metric and one of the good indicators of damage of gears. Kurtosis can be estimated in both the time domain and the frequency domain. This paper discusses various variants of the kurtosis parameter such as FM4, NA4, NA4 * , NB4, and NB4 * , as well as their relative importance. The ability to use kurtosis beta distribution to detect toothwise gear faults is also discussed.