Deep Recurrent Entropy Adaptive Model for System Reliability Monitoring

The aim of this article is to develop a methodology for measuring the degree unpredictability in dynamical systems with memory, i.e., responses dependent on history past states. proposed model generic, and can be employed variety settings, although its applicability here examined particular context an industrial environment: gas turbine engines. given approach consists approximating probability distribution outputs system deep recurrent neural network; such networks are capable exploiting memory enhanced forecasting capability. Once retrieved, xmlns:xlink="http://www.w3.org/1999/xlink">entropy or xmlns:xlink="http://www.w3.org/1999/xlink">missing information about underlying process computed, which interpreted as uncertainty respect system's behavior. Hence, identifies how far dynamics from typical response, order evaluate reliability predict faults and/or xmlns:xlink="http://www.w3.org/1999/xlink">normal accidents . validity verified sensor data recorded commissioning turbines, belonging normal faulty conditions.