Songpon Klinchaeam, Nopdanai Ajavakom, and Withaya Yongchareon

pp. 563 - 568

Abstract

This paper presents a condition monitoring technique of a spur gear fault detection using vibration signal analysis
based on time domain. Vibration signals were acquired from gearboxes and used to simulate various faults on spur gear
tooth. In this study, vibration signals were applied to monitor a normal and various fault conditions of a spur gear such as
normal, scuffing defect, crack defect and broken tooth. The statistical parameters of vibration signal were used to compare
and evaluate the value of fault condition. This technique can be applied to set alarm limit of the signal condition based on
statistical parameter such as variance, kurtosis, rms and crest factor. These parameters can be used to set as a boundary
decision of signal condition. From the results, the vibration signal analysis with single statistical parameter is unclear to
predict fault of the spur gears. The using at least two statistical parameters can be clearly used to separate in every case of
fault detection. The boundary decision of statistical parameter with the 99.7% certainty ( ±3σ )  from 300 referenced dataset
and detected the testing condition with 99.7%  ( ±3σ ) accuracy and had an error of less than 0.3 % using 50 testing dataset.