Memristor-based LSTM neuromorphic circuits for offshore wind turbine blade fault detection

The UK offshore wind industry is rapidly growing to meet CO2 emission targets. However, the main drawback of the offshore environment is the increased cost of maintenance. Artificial Neural Networks (ANN) show great potential to reduce this cost. Long Short-Term Memory (LSTM) is a form of Recurrent Neural Network (RNN) that shows promising results in solving time series-based problems, making them ideally suited for wind turbine condition monitoring. A dedicated circuit for a LSTM-based ANN that uses memristors will allow for more power efficient and faster computation, whilst being able to overcome the von Neumann bottleneck.