The application of Bayesian inference to RUL predictions of the tool was demonstrated using a random walk approach, where the prior probability of FMWwas generated using sample FWW growth curves that represented the true FWW growth curve with some probability. This probability was updated using Bayesian inference. Although a linear FWW growth model was assumed in this study, a higher order model may also be assumed to describe the three stages of tool wear [3]. The method can be extended to include sensor data such as power or acoustic emission. In addition, uncertainty regarding the threshold value of the sensor, such as percent increase from the nominal, can also be incorporated.
展开▼