Normal view MARC view ISBD view

Classification of electronic devices using harmonic radar based on a linear model with power-swept signals / Maryam Shahi ; thesis advisor Harun Taha Hayvacı.

By: Shahi, Maryam [author].
Contributor(s): Hayvacı, Harun Taha [advisor] | TOBB Ekonomi ve Teknoloji Üniversitesi Fen Bilimleri Enstitüsü Fen Bilimleri Enstitüsü.
Material type: materialTypeLabelBookPublisher: Ankara : TOBB ETÜ Fen Bilimleri Enstitüsü, 2019Description: ix, 35 pages : illustrations ; 29 cm.Content type: text Media type: unmediated Carrier type: volumeSubject(s): Tezler, Akademik | Harmonic radar | Power series model | Maximum likelihood estimator | Linear model | K-nearest neighbors | Classification | Vectors of parametersOnline resources: Ulusal Tez Merkezi Dissertation note: Tez (Yüksek Lisans)--TOBB ETÜ Fen Bilimleri Enstitüsü Temmuz 2019 Summary: Nonlinear circuit components such as diodes, transistors, etc. receive a transmitted signal at a fundamental frequency and re-radiate the harmonics and possibly intermodulation products of that frequency. Many studies have been done to utilize nonlinearities of electronic circuits to detect, range, and track targets of interest in the presence of clutter using harmonic radar, yet only a few number of researches focus on classification of various electronic circuits. A new technique to use nonlinear characteristics of electronic devices for classification of those devices using Harmonic Radar is proposed in this thesis. Unlike prior work in the literature, the powers of the transmitted incident waves are swept within a determined range in this study to capture the nonlinearities of the Electronic Circuits Under Test (ECUT). National Instruments (NI) AWR Design Environment is used to design the ECUT of this study. The first three harmonics of the received powers are analyzed in harmonic space. This novel method, derives the harmonic responses of the ECUT using Power Series Model. As a major contribution of this research, a linear model is proposed to relate the measurements to the unknown deterministic vectors of parameters characterizing the nonlinearities of the ECUT. Each electronic circuit under test has a distinguishable harmonic response to a single-tone or two-tone incident wave with varying power. Therefore, a unique vector of parameters can be derived from the presented linear model for each circuit. A Maximum Likelihood Estimator (MLE) is used in this novel approach to estimate the unique vectors of parameters in the presence of Complex White Gaussian Noise (CWGN) based on the newly developed linear model. K-Nearest Neighbors (kNN) classification method is employed to classify different nonlinear electronic devices such as diode clamper, diode limiter, and full-wave rectifier using the statistical features of the normalized estimated vectors of parameters as distinguishing factors. Simulation results prove the presented method of power-swept incident waves and estimated vectors of parameters to be an effective approach for classification of nonlinear devices using harmonic radar. The performance of the obtained classifier is evaluated using confusion matrices and scattered feature plots of the normalized estimated vectors of parameters. It is shown that the presented classifier in this study has a better performance compared to existing classifiers.
Tags from this library: No tags from this library for this title. Log in to add tags.
    average rating: 0.0 (0 votes)
Item type Current location Collection Call number Copy number Status Date due Barcode
Thesis Thesis Merkez Kütüphane
Tez Koleksiyonu / Thesis Collection
Tezler TEZ TOBB FBE ELE YL’19 SHA (Browse shelf) 1 Ödünç Verilemez-Tez / Not For Loan-Thesis TZ01024

Tez (Yüksek Lisans)--TOBB ETÜ Fen Bilimleri Enstitüsü Temmuz 2019

Nonlinear circuit components such as diodes, transistors, etc. receive a transmitted signal at a fundamental frequency and re-radiate the harmonics and possibly intermodulation products of that frequency. Many studies have been done to utilize nonlinearities of electronic circuits to detect, range, and track targets of interest in the presence of clutter using harmonic radar, yet only a few number of researches focus on classification of various electronic circuits. A new technique to use nonlinear characteristics of electronic devices for classification of those devices using Harmonic Radar is proposed in this thesis. Unlike prior work in the literature, the powers of the transmitted incident waves are swept within a determined range in this study to capture the nonlinearities of the Electronic Circuits Under Test (ECUT). National Instruments (NI) AWR Design Environment is used to design the ECUT of this study. The first three harmonics of the received powers are analyzed in harmonic space. This novel method, derives the harmonic responses of the ECUT using Power Series Model. As a major contribution of this research, a linear model is proposed to relate the measurements to the unknown deterministic vectors of parameters characterizing the nonlinearities of the ECUT. Each electronic circuit under test has a distinguishable harmonic response to a single-tone or two-tone incident wave with varying power. Therefore, a unique vector of parameters can be derived from the presented linear model for each circuit. A Maximum Likelihood Estimator (MLE) is used in this novel approach to estimate the unique vectors of parameters in the presence of Complex White Gaussian Noise (CWGN) based on the newly developed linear model. K-Nearest Neighbors (kNN) classification method is employed to classify different nonlinear electronic devices such as diode clamper, diode limiter, and full-wave rectifier using the statistical features of the normalized estimated vectors of parameters as distinguishing factors. Simulation results prove the presented method of power-swept incident waves and estimated vectors of parameters to be an effective approach for classification of nonlinear devices using harmonic radar. The performance of the obtained classifier is evaluated using confusion matrices and scattered feature plots of the normalized estimated vectors of parameters. It is shown that the presented classifier in this study has a better performance compared to existing classifiers.

This software was implemented, installed by Devinim Software Training Consulting .