High Performance Remaining Useful Life Prediction for Gearbox

Gearbox failure prediction is important for many applications. We present four approaches for predicting the remaining useful life of gearboxes. The first one is rule-based and the second one is a combination of damage curve approach and rule-based approach. The third one is using Quadratic Discriminant Analysis and the final one is a combination of the first three. Gearbox data from the 2008 Prognostic and Health Management (PHM) conference challenge were used to evaluate our algorithms. The results are close to those top performers in the PHM Challenge.