The forecasting research on the model of the hysteresis brake based on Elman Multi-Hidden Layers Neural Network algorithm with the speed and directional factor

Hysteresis brake, which provides for step-less torque and high speed, is a stable and reliable device used for loading, but with some problems, such as the large time-lag, strong nonlinearity, and difficulty in modeling et al. Due to the above-mentioned issues, the special working conditions of hysteresis brake developed by our project team and the advantage of the dynamical performance of the Elman Multi-Hidden Layers Neural Network, this paper employs modified neural network algorithm with Elman dynamic recursive structure to accomplish the model of hysteresis brake, and guarantees the accuracy of output loaded torque. In this algorithm, we introduce the speed, the given and desired torque, and a directivity factor respectively as a disturbance variable and two input vectors of the system, and then train the model and test. The numerical results validate the effectiveness of the algorithm proposed.