Abstract: With the broad deployment of sensors in industrial equipment, data-driven device state prognostics and health management have received increasing attention from both academia and industry. This paper focused on prognostics of systems remaining useful life (RUL). Deep neural networks to build the key step of RUL estimation models. We evaluated the RUL estimation performance of the models using three typical deep neural networks, namely, feed-forward neural network (FNN), convolution neural network (CNN), and long and short-term memory (LSTM), based on a benchmark dataset C-MAPSS. The experimental results demonstrate that LSTM considering temporal features have significant performance advantages. The research trends in RUL prediction are discussed.