Design And Implement Of Radio Frequency And Microwave Devices Modeling And Simulation Based On Neural Networks

With the rapid development of Electronic Science and Technology, RF and microwavecircuits have been widely used in various related fields, such as mobile communications, radar,electronic warfare technology and instrumentation, and have reached a wide range ofdevelopment. Accurate design and analysis of RF and microwave circuit models play a crucialrole in its application, so people committed to looking for more quickly and accurately modeldesign and analysis methods. On the basis of weakly nonlinear, traditional design and analysismethods are no longer applicable to the strongly nonlinear microwave circuits, therefore,urgent need to open up a new theory to solve this problem. The neural network technologyassisted microwave circuit modeling method is an effective method to choose.This paper first describes the basic of artificial neural networks and its application in RFand microwave circuit modeling, and further proposed neural network-assisted RF andmicrowave circuit modeling method, the neural network directed modeling method andknowledge-based modeling method. Compared to traditional modeling methods, this neuralnetwork-assisted modeling approach can achieve high design accuracy with very littlecomputational. The paper also introduces the artificial neural network modeling andsimulation tools, NeuroModeler and NeuroADS. Developed neural network modelssuccessfully for RF and microwave passive components using the direct modeling approach,such as the microstrip stepped impedance filter, rectangular microstrip patch antenna and themicrostrip hairpin filter. Within the scope of the neural model’s learning and training,researchers can arbitrarily change the value of the variable according to design requirementsand can obtain the electromagnetic response quickly and accurately, significantly reduce theCPU time consuming.Meanwhile, on the basis of the neural network direct modeling for RF and microwavepassive components, the paper also proposed a new knowledge-based neural networkmodeling approach，the high-dimensional neural network modeling techniques. This modelcan be used to solve the issue of multi-variable RF and microwave passive device modeling,which avoids microwave complex mathematical operations needed in the analysis ofequivalent circuit of the sub-module of the nonlinear circuit. Relatively speaking, thisknowledge-based modeling of high-dimensional neural network model can save a lot of timefor the entire modeling process compares to the traditional EM simulation and neural networkdirect modeling approach. And the use of this method to design a microstrip hairpin filter,verify the accuracy of the theory.