High Order Narrow Band Superconducting Filter Design Based on Neural Networks and Extracted Coupling Matrix

In this article, an effective high-order superconducting filter design technique is proposed. The key advancement is the demonstration of neural network high temperature superconducting (HTS) high-order filter design for the first time. An artificial neural network (ANN) combined with the coupling matrix is used as a fast model of the high-order superconducting filters. The initial filter layout is established and characterized by a set of geometric variables. The coupling matrix is extracted from the simulated response of the filter layout, and the ANN is trained to learn the relationship between the coupling matrix and the filter geometric variables. The well-trained neural network model can provide an accurate and fast prediction of filter performance with different input geometric variables in less than one second, which greatly improve the design effectiveness. The experiments show that there is an excellent match between the responses of the simulated data and those from the neural network. Compared with the conventional electromagnetic simulation method, this model is time saving especially in high-order filters design with complex undesired stray couplings.