Aristotle University Of Thessaloniki (AUTH)
Μέθοδοι εξέτασης βλαβών αναλογικών και μικτών κυκλωμάτων
Abstract
dc:descriptionIn this PhD thesis methods for circuit fault detection were investigated. Three fault detection methods were developed, one that is based on the spectrum of the supply current of the circuit, one that is based on the wavelet transform and one that is based on the Support Vector Machines which is a technique that is used in pattern recognition. Furthermore a theory for analog test metric estimation is first time demonstrated in a realistic case study. At first one fault detection method which is based in the Fourier transform of the supply current was developed in order to be utilized in real circuits. The term reference circuit was introduced which replaces the term nominal circuit. The values of the reference circuits are calculated by the mean value of the fault free circuits. The method was utilized in order to test an emergency luminaire circuit which is a commercial product and a continuous time state variable filter (Benchmark circuit from the ITC97 suite) which was implemented using discrete components. The next method is based on the wavelet transform. The metric that was used is the energy of the trend and the detail coefficient of the first level of decomposition. In comparison with other methods in the literature, this method has the advantage to use only one value to discriminate the faulty from the fault free circuits. This method can be applied using only the energy of the trend coefficients or only the energy of the detail coefficients or its combination (according to the circuit under test). This method is appropriate to be used in a microcontroller based system, because the implementation of the wavelet transform is just the implementation of an FIR digital filter bank. The efficiency of the method was verified using the emergency luminaire circuit and the continuous time state variable filter. In the second method which uses the wavelet transform two new test metrics were introduced. The first uses the energy of the trend coefficients of the second level of decomposition and the detail coefficients of the first level of decomposition. Actually this metric is a normalized Euclidian distance. The second metric is based on the Mahalanobis distance. The Mahalanobis distance is a statistical distance which takes into account the correlation of the data. This metric was utilized using the energy of the trend and the detail coefficient of the first level of decomposition. The tolerance limits for those methods are calculated from the cumulative probability functions of the fault free circuits. This method was used in order to detect parametric and catastrophic faults. The circuits in which were used are the emergency luminaire circuit, the continuous time state variable filter, the leapfrog filter and an operational amplifier LM741. The impact of the mother function of the wavelet transform in the fault coverage as long as the sensitivity of the method in other fault models were investigated. The third method that was developed is based on the Support Vector Machines. Two different classes of data were considered, the faulty and the fault free circuits. The SVM model was trained with two different kinds of features. In the first case the model was trained using magnitude of the harmonics of the spectrum and the RMS value of the supply current. In the second case the model was trained using the energy of the coefficients of the wavelet transform of the supply current.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2012
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Spyronasios, Alexios
- Σπυρονάσιος, Αλέξιος
Subjects
dc:subject × 12- Μετασχηματισμός κυματιδίων
- Εξέταση βλαβών
- Έλεγχος κυκλωμάτων
- Όρια διαχωρισμού
- Fault detection
- Wavelet transform
- Circuit testing
- Torelance limits
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Engineering and Technology
- Electrical Engineering, Electronic Engineering, Information Engineering
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/29132