Role of artificial neural networks and wavelets in online reliability monitoring of physical systems

In some cases, degradation signals exhibited by a physical component contain valuable information regarding the state or well-being of the component. We introduce several methods that facilitate estimation of individual component reliability utilizing online degradation signals. Even though degradation signals can be modeled using different tools, the paper discusses the relevance of neural networks for modeling the same. In the case when the waveform or signature characteristics of the signal are more important than the amplitude characteristics, the paper recommends a wavelet transform prior to degradation signal modeling. The proposed techniques have been applied to monitor drill bits used in a machining center. The degradation signals, force and torque, were collected as the drill bits were destructively tested