Deep residual ensemble model for predicting remaining useful life of turbo fan engines

Abstract Capturing degradation trends from the Condition monitored signals is a proven technique for predicting the Remining Useful Life (RUL) of the equipment, which has gained more prominence in Prognostics and Health Management (PHM) in Industry 4.0. However, this process is tiresome and comprehending all the physical parameters of the system to construct a Health Index that characterize the health state is a complex process, especially if multiple sensors are involved. This work proposes a Deep residual ensemble model which constructs Fused Health Index (FHI) by harnessing temporal property of signals. The proposed Residual network integrates Bi-directional Long Short Term Memory (Bi-LSTM) and Deep Neural Network (DNN) which absorbs individual residuals of both the forward and reverse LSTMs that acts as an important feature to improve the overall prediction process. The work validated using CMAPPS dataset using various unique performance metrics to portray the effectiveness of the model.