This study demonstrates how a time domain data based non-linear approach known as Stochastic Resonance (SR) can be effectively used for fault detection in spur gearboxes. SR has just been used recently for fault diagnosis in mechanical systems with a focus on faulty systems. This paper examines the behavior of SR when it is applied to healthy systems, in particular a healthy gearbox and explores approaches like residual signal and filtered signal computations to aid in the containment of false alarms while improving overall results. Although SR is a time domain procedure, its results also extend to the frequency domain.