A new approach for gear wear state identification based onHidden semi-Markov Model which is represented as Dynamic Bayesian Networks is presented.One can recognize the gear wear state through identifying the HSMM-DBN that best fits the observations.According to the non-linear dependencies between features

the curvilinear distance analysis is used for dimension reduction.Finally

the data of gear wear tests are used to demonstrate the proposed methods.The results show that the gearwear state can be effectively identified and the accuracy can be reach 94.5%

a scientific basis for the health management of gearbox is provided.