Remaining Useful Life Prognosis for Rolling Bearings Based on Bi-ConvLSTM

The construction of health indicators (HI) is an essential task for RUL prognosis. However, the traditional HI construction methods rely on the prior knowledge of experts, thus seriously affecting the prognosis accuracy of RUL. In this paper, a Bi-ConvLSTM model is established to adaptively construct a HI corresponding to the degree of performance degradation. Combined with the dual index prognosis model and particle filter algorithm, it is applied to the RUL prognosis of bearings. Experimental results show that the RUL prognosis accuracy of Bi-ConvLSTM is higher than that of traditional RUL prognosis frameworks.