A hybrid evolutionary cnn-lstm model for prognostics of c-mapss aircraft dataset
Abstract
Summary & ConclusionsThe fundamental concept of prognostics and health management (PHM) is to find an approach to evaluate the system's health and predict its remaining useful life (RUL). In the era of digital transformation, many methods and algorithms from the data science world have been applied to use for PHM modeling. One of the leading algorithms is deep learning. Many of the PHM models proposed during the past few years have shown a significant increasing trend of employing hybrid deep neural network schemes. The hybrid modeling approach is using the combination of multiple types of neural network layers to construct the model architecture which can often provide more accuracy but rather too complex. One possible approach to reduce the complexity of the model is to use the features selection methods prior to the model training process. In this work, the evolutionary selection method was used to select features from C-MAPSS aircraft gas turbine engine dataset [1]. The features selected from evolutionary selection, were then, used to train a hybrid Convolutional Long Short-Term Memory (CNN-LSTM) deep neural network for the C-MAPSS RUL prediction model. The C-MAPSS dataset was derived from the NASA Ames prognostics data repository. The findings show that applying evolutionary selection, combined with CNN-LSTM helps to improve the overall PHM model's performance in both complexity and accuracy.
Authors: Phattara Khumprom, Alex Davila‐Frias, David Grewell, Dollaya Buakum