IMMG Student Projects

2002-2003 Graduate Project

Radial-Basis-Function Neural Network Optimization of Microwave Systems

Student: Ethan Murphy, GS, '03
Advisor: Vadim Yakovlev

Abstract

An original approach in microwave optimization, namely, a neural network procedure combined with the full-wave 3D electromagnetic simulator QuickWave-3D implemented a conformal FDTD method, is presented. The radial-basis-function network is trained by simulated frequency characteristics of S-parameters and geometric data of the corresponding system. High accuracy and computational efficiency of the procedure is illustrated for a waveguide bend, waveguide T-junction with a post, and a slotted waveguide as a radiating element.

Contents of the Report

Chapter 1. Introduction
Chapter 2. Background
2.1 Electromagnetic Issues
2.2 Basics of Neural Networks
2.3 Neural Networks in Microwave Modeling
Chapter 3. Analysis
3.1 Statement of the Problem
3.2 Feedforward MLP NN
3.3 RBF NN
3.4 Optimization Method
Chapter 4. Neural Model
Chapter 5. Implementation
5.1 Overview
5.2 Creation of the Database
5.3 Construction and Training of Radial Basis Network
5.2 Optimization and Comparison
Chapter 6. Illustrations
6.1 MW Systems
6.2 Scaling
6.3 Accuracy
6.4 Optimization
6.5 Comparison with QuickWave-3D's Optimizers
Chapter 7. Conclusions
Appendix
Bibliography

Motivations and Goals

The modern trends towards production-oriented design and reduced time-to-market in the microwave industry require instruments assisting in accurate and fast design. Efforts to lower the cost and reduce the weight/volume of the circuits have caused a keen interest of electronic and microwave engineers in new efficient CAD tools.

However, a simple application of highly sophisticated computational tools for analysis of microwave systems may not bring many direct recommendations for design implementation. Practical problems may be associated with specific optimization goals, which cannot be addressed with the use of the general tools in the software packages. This dictates the necessity of development of efficient optimization techniques for microwave modeling. Efficient computational procedures linked with advanced EM solvers should become powerful and flexible CAD tools revolutionizing the design of microwave devices.

For the first time, the present paper proposes an efficient and simple optimization technique based on artificial neural networks (NN) made as a computational supplement for the 3D conformal FDTD simulator QuickWave-3D. It is shown that given the resources of today's computers such an approach can be reasonably productive and serve as a competent optimization tool in designing of various MW systems.

Selected Results

A c c u r a c y

Geometry of the 11.5 x 23 mm waveguide bend (left) and the magnitude of the reflection coefficient in this bend for p = 0, f = 10.5 GHz and varying m and s (in mm): accurate computation by QuickWave-3D (center) and the result generated by the developed RBF NN with a 125-sample database(right)

O p t i m i z a t i o n

Example of 3-parameter (h, r, s) minimization of |S11| in the frequency range (f1, f2). Geometry of the WR75 (19.05 x 9.53 mm) T-junction with a post (left) and optimized frequency characteristic |S11| in this junction: obtained by the RSM-SQP-method (green curve) and by the developed RBF NN technique with the use of databases of different sizes (right).

Example of 3-parameter (s, d, Theta) minimization of |S11| in the frequency range (f1, f2). Geometry of the slotted waveguide-backed antenna (left) and the optimized frequency characteristic of |S11| in this structure for WR430 (86 x 43 mm), w = 8 mm, and l = 65 mm obtained by the RSM-SQP-method (green curve) and by the developed RBF NN technique with the use of databases of different sizes (right).

vadim@wpi.edu

Last modified: Tue, Aug 16, 2005