Aircraft equipment modeling using neural networks

In this paper, the neural network module for simulating the behavior of an arbitrary system is described. The black-box modeling tool is aimed to simulate unknown systems characterized by measurements. The tool is able to model systems that cannot be described analytically due to their nonlinearity or complexity. Neural networks are suitable for solving real EMC problems leading to extremely high computation demands or to complicated experimental measurements. There are two main limitations: the sparsity of numerical results (due to the computation complexity) and the noise corrupting the measurement results. It is essential to find an equivalent model approximating the behavior of investigated system if the sparse or noisy data are available.