A Bidirectional GRU and CNN-Based Deep Learning Method with Optimized Structure by Genetic Algorithm for Predicting Remaining Useful Life of Turbofan Engines

Accurate predictions of the remaining useful life (RUL) of turbofan engine plays an important role in system reliability, which is the basis of prognostics and health management (PHM). this paper proposes a hybrid deep learning method consisting of a convolutional neural network (CNN) and a bidirectional gated recurrent unit (BiGRU) called the CNN-BiGRU hybrid to improve predictive performance. This hybrid structure also has extensive hyperparameters that not only affect the accuracy of model but also affect the selection of some other hyperparameters, so the genetic algorithm is applied to obtain the optimal hyperparameters of the CNN_BiGRU structure. The effectiveness of the proposed design is confirmed on NASA Commercial Modular Propulsion Aircraft Simulation Database (C-MAPSS). The proposed prediction method for this multivariate time series dataset works better than the previous methods based on this dataset.

A Bidirectional GRU and CNN-Based Deep Learning Method with Optimized Structure by Genetic Algorithm for Predicting Remaining Useful Life of Turbofan Engines Keywords:

Hybrid Neural Network , Multivariate time series forecasting , Genetic algorithm , Remaining useful life