A neural network method for the analysis of multilayered shielded microwave circuits

This paper proposes the design and analysis of practical multilayered shielded microwave circuits using neural networks. The radial basis function neural network (RBFNN) approximates, with the accuracy demanded, the boxed Green´s functions needed in the integral equation (IE) method. After the initial neural network training, the RBFNN outputs substitutes the exact values of the Green´s functions, leading to a large reduction in the calculation time of the essential circuit parameters. In this paper, a novel strategy based on spatial and frequency regions subdivisions is presented. This strategy allows to reduce the computational burden associated to the training of the neural network. The computational gain obtained with the neural network method allows the development of neural based computer-aided-design (CAD) tools, that can be used for the analysis and design of shielded printed MMIC devices in a near real time basis.