Prognostics in Aeronautics with Deep Recurrent Neural Networks

摘要: Recurrent neural networks (RNNs) such as LSTM and GRU are not new to the field of prognostics. However, performance strongly depends on their architectural structure. In this work, we investigate a hybrid network architecture that is combination recurrent feed-forward (conditional) layers. Two networks, one another feed-forward, chained together, with inference weight gradients being learned using standard back-propagation learning procedure. To better tune network, instead raw sensor data, do some preprocessing mostly simple but effective statistics (researched in previous work). This helps feature extraction phase eases problem finding suitable configuration among immense set possible ones. first proposal prognostics our work novel sense it performs more comprehensive comparison type for different RNN layers number Also, compare other classical machine methods. Evaluation performed two real-world case studies from aero-engine industry: involving critical valve subsystem jet engine whole reliability engine. Our goal here cases contrasting micro (valve) macro (whole engine) results indicate deep significantly than models.