Modelling hysteresis with neural network

An example of a system with hysteresis is an actuator made of SMA. In the paper, the LSTM neural network was used to model the actuator. The focus was pointed to its hysteretic part. Neural network structure and learning method are briefly presented. Data used for training was obtained from the Preisach hysteresis model. This model was used to generate the datasets used for training and testing. Thanks to this approach, we can generate data quickly and in a fully controllable way. Thanks to this, experiments can be thoroughly planned and fully repeatable. The advantage and disadvantage at the same time is the lack of disturbances. The main goal of the paper was to model hysteresis on an example of an SMA actuator, but in fact, it was interesting if a neural network could describe hysteresis loop because literature research shows that usually neural networks are used to model hysteresis with a hysteretic element in combination modelled in another way.