Modeling of Spiral Inductors Using Artificial Neural Network

T Liu,WJ Zhang,ZP Yu
DOI: https://doi.org/10.1109/ijcnn.2005.1556269
2006-01-01
Abstract:A new model for spiral inductors, which covers wide operation frequency range and full design parameters, is proposed by using artificial neural network (ANN). It is pointed out that a four-layered neural network is superior to a three-layered neural network both on the mapping and generalization abilities in spiral inductor modeling. For the first time, a novel physics-based sampling technique is adopted in modeling procedure. Equipped with this new sampling method, ANN model achieves better speed and accuracy performances though the training data are substantially reduced. The new sampling method can be easily applied to other passive components and be embedded in various modeling frameworks.
What problem does this paper attempt to address?