A Novel Electromagnetic Centric Multiphysics Parametric Modeling Approach Using Neuro-Space Mapping for Microwave Passive Components

Self Cite

An analytically separated neuro-space mapping (Neuro-SM) model of power transistors is proposed in this paper. Two separated mapping networks are introduced into the new model to improve the characteristics of the DC and AC, avoiding interference of the internal parameters in neural networks. Novel analytical formulations are derived to develop effective combinations between the mapping networks and the coarse model. In addition, an advanced training approach with simple sensitivity analysis expressions is proposed to accelerate the optimization process. The flexible transformation of terminal signals in the proposed model allows existing models to exceed their current capabilities, addressing accuracy limitations. The modeling experiment for the measurement data of laterally diffused metal-oxide-semiconductor transistors demonstrates that the novel method accurately represents the characteristics of the DC and AC of transistors with a simple structure and efficient training process.

Section: Introductionmentioning

confidence: 99%

Analytical Separated Neuro-Space Mapping Modeling Method of Power Transistor

Wang

Yan

et al. 2023

Self Cite

“…Space mapping (SM) [ 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 ] technology has been widely used to solve the growing computational challenges in the field of microwave modeling. The SM algorithm assumes the existence of coarse models [ 5 ], and uses empirical functions [ 15 ] or equivalent circuits [ 16 , 17 , 18 ] as prior knowledge to assist in the modeling process of microwave devices. The SM combines the high computational efficiency of the coarse model and the high fidelity of the EM fine model [ 19 ].…”

Section: Introductionmentioning

confidence: 99%

A Mesh Space Mapping Modeling Method with Mesh Deformation for Microwave Components

Yan,

Li,

et al. 2023

Self Cite

In this study, a low-cost space mapping (SM) modeling method with mesh deformation is proposed for microwave components. In this approach, the coarse-mesh model with mesh deformation is developed as the coarse model, and the fine-mesh model is simulated as the fine model. The SM technique establishes the mapping relationship between the coarse-mesh model and the fine-mesh model. This approach enables us to combine the computational efficiency of the coarse model with the accuracy of the fine model. The automatic mesh deformation technology is embedded in the coarse model to avoid the discontinuous change in the electromagnetic response. The proposed model consisting of the coarse model and two mapping modules can represent the features of the fine model more accurately, and predict the electromagnetic response of microwave components quickly. The proposed mesh SM modeling technique is applied to the four-pole waveguide filter. The value for the training and test errors in the proposed model is less than 1%, which is lower than that for the ANN models and the existing SM models trained with the same data. Compared with HFSS software, the proposed model can save about 70% CPU time in predicting a set of 100 data. The results show that the proposed method achieves a good modeling accuracy and efficiency with few training data and a low computational cost.

“…By adjusting the MNNs, the coarse model is gradually adapted to the characteristics of the fine model, enabling the Neuro-SM model to achieve both high accuracy and high simulation efficiency. This method has been widely used for device modeling in the microwave field [ 29 , 30 , 31 , 32 , 33 , 34 ].…”

Section: Introductionmentioning

confidence: 99%

“…The frequency MNNs added to the Neuro-SM model can enhance the frequency characteristics of the coarse model [ 36 ]. In [ 29 ], the coarse model introduces both the input and output MNNs in order to obtain a more accurate model. If there is a significant difference between the fine model and the empirical formulas of the coarse model, the existing Neuro-SM modeling methods cannot develop a precise model.…”

Section: Introductionmentioning

confidence: 99%

Improved Empirical Formula Modeling Method Using Neuro-Space Mapping for Coupled Microstrip Lines

Yan

Qian

et al. 2023

Self Cite

In this paper, an improved empirical formula modeling method using neuro-space mapping (Neuro-SM) for coupled microstrip lines is proposed. Empirical formulas with correction values are used for the coarse model, avoiding a slow trial-and-error process. The proposed model uses mapping neural networks (MNNs), including both geometric variables and frequency variables to improve accuracy with fewer variables. Additionally, an advanced method incorporating simple sensitivity analysis expressions into the training process is proposed to accelerate the optimization process. The experimental results show that the proposed model with its simple structure and an effective training process can accurately reflect the performance of coupled microstrip lines. The proposed model is more compatible than models in existing simulation software.