Applying artificial neural networks in automatic control systems for magnetic levitation

This paper considers the problem of neural network identification in the automatic control system for magnetic levitation (neural network controller). This problem is due to the fact that at present, it is typical to use a mathematical apparatus based on artificial neural networks when designing automatic control systems for complex technical dynamic objects. Such networks have the following unique advantages: the possibility of parallel computing; finding previously unknown relationships between in-put and output sequences of digital signals; more efficient control of nonlinear systems using nonlinear activation functions. In addition, the use of artificial neural networks sometimes removes the difficulties that arise when describing some problems in the form of analytical mathematical models. This paper describes the solution for the problem based on artificial intelligence methods – well-known recurrent artificial neural networks NARX and LSTM. These networks were trained using the backpropagation algorithm and the Levenberg–Marquardt method, which have good convergence. For this problem, the application of trained artificial neural networks to test data has shown some advantage of the NARX network compared to LSTM. Moreover, the RSME rms error for the NARX network with 50 hidden layers for this problem is less than for the network with 100 hidden layers. The MATLAB R2021b system was a tool for setting the architecture of neural networks, their construction, training and testing. The obtained results are applied in designing automatic control systems for levitating objects. Currently, most of these systems are being developed and used to transport goods for various purposes using the effect of magnetic levitation.