Modeling of X-Y macro-positioning stage based on non-smooth neural networks

In this paper, a novel neutral-network-based model is proposed to describe an X-Y macro-positioning stage. As the friction exists in the stage, the stage shows some complex behavior due to the non-smooth characteristic of the friction. In order to describe the non-smooth behavior of the stage, in this model, a non-smooth active function is proposed to construct the hidden neurons. Then, a training algorithm cooperated with the generalized gradient technique is developed to train the proposed neural network. Finally, the experimental results are presented to illustrate the performance of the proposed method.