A Remaining Useful Life Prediction Framework for Multi-sensor System

As the key technology of prognostics and health management, the remaining useful life (RUL) prediction can effectively reduce the fault probability and maintenance cost by evaluating the system operation condition. Due to the complex structure and versatility, an engineering system requires multiple sensors to monitor its condition. Therefore, developing methodologies capable of integrating data from multiple sensors is important for accurate RUL prediction of multi-sensor system. In this paper, an RUL prediction framework for multi-sensor system is put forward by combining data fusion and particle filter. Specifically, permutation entropy is introduced to describe degradation trend and a composite health index is constructed through the fusion of multi-sensor data. By constructing the deterioration model of composite health index, the particle filter algorithm is used to realize RUL prediction for multi-sensor system. A set of degradation data of an aircraft gas turbine engine from NASA Ames Research Center is used to validate the effectiveness of proposed prediction framework.