Neural network with knowledge‐based neurons for the modeling of crossover discontinuities in stripline circuits

Abstract

ABSTRACT:A novel three-layer neural network with knowledge-based neurons in hidden layer (NNKBN)

Neural Network Model with Broadband Prior Knowledge Neurons for Microstrip T-junction Structure

“…3. Fig.3 The equivalent lumped parameters circuit for microstrip T-junction structure According to the establish method of NNKBN model [1][2], the broadband characteristics prior knowledge is utilized to substitute the prior knowledge in neurons of hidden layer of NNKBN model, thus we can establish a novel KBNN model. The ranges and sampling levels of the input variables x are as given in Table I.…”

“…NNKBN model had been widely used for modeling many stripline discontinuities and microstrip discontinuities [1][2], NNKBN is proved to have more accurate and more reliable extrapolate ability than conventional MLP model. However, there are exists two serious problems that limited the practical use scope of NNKBN model, one is these classic analytic equivalent circuit equations used in NNKBN model lack of broadband characteristic so that it's necessary to obtain much training data in a wide frequency band, the robust ability of neural network need further improvement.…”

Robust knowledge‐based neural‐network model for microstrip T‐junction structure

“…A popular method is the use of parasitic patches, either in another layer (stacked geometry) [4] or in the same layer (coplanar geometry). However, the stacked geometry has the disadvantage of increasing the thickness of the antenna, and the coplanar geometry has the disadvantage of increasing the lateral size of the antenna [5]. In a conventional microstripline-fed slot antenna, the narrow rectangular slot is cut in the ground plane, and the slot is excited by a microstrip feed line with a short [6] or an open [7] termination.…”

“…In the last few years, multilayer perception (MLP) neural networks have been used for the modeling of discontinuities in stripline circuits [3][4][5]. A potential drawback of the MLP models is the huge amount of training data needed for obtaining an accurate model.…”