Development of Recurrence Analysis for Fault Discrimination in Gears

Most of the current approaches in gear fault diagnostic analysis do not provide a clear indication for discrimination between various common faults. This is almost certainly because failure in gear systems is always accompanied with a complex dynamic response. The inherent nonlinearities in geared systems and additional nonlinearity due to faults conspire to make the diagnostics problem extremely complex. The main goal of this paper is to exploit the nonlinear response in an attempt to solve this challenging problem. We aim to demonstrate effective discrimination between four kinds of common defects in gears by utilizing the method of recurrence analysis. We consider a healthy system, a system with a single crack, a system with multiple cracks and a system with a missing tooth. Careful measurements are collected from an experimental set up which is a mock-up of a real industrial gear box system. We develop and apply recurrence plot analysis to this data. Our results show that recurrence analysis parameters such as recurrence rate, trapping time, entropy and maximal diagonal length can be very effective in successful discrimination of gear faults.