EPE Association: EPE 1997 – 49: Dialogue Session DS9f: DIAGNOSTICS

EPE 1997 – 49: Dialogue Session DS9f: DIAGNOSTICS
You are here: EPE Documents > 01 - EPE & EPE ECCE Conference Proceedings > EPE 1997 - Conference > EPE 1997 – 49: Dialogue Session DS9f: DIAGNOSTICS

   [return to parent folder]

   USE OF THE OUTPUT VOLTAGE OF A SWITCHMODE POWER SUPPLY TO PREDICT FAILURE OF FILTERING ELECTROLYTIC CAPACITORS
By P. Venet; A. Lahyani; P.J. Viverge; G. Grellet [View]
[Download]

Abstract: Filtering electrolytic capacitors are responsible of most breakdown cases in static converters. The best indicator of fault of these capacitors is the equivalent series resistance (ESR) which increases versus temperature and ageing time according to a well known law. Since the output voltage ripple of the converter is almost proportional to ESR, the failure of the capacitors is predicted by monitoring online this voltage during operation of the converter. So as to avoid errors due to sudden load variation, the output voltage ripple is firstly filtered at the switching frequency of the converter, then processed with other converter parameters, thereby, the lifetime of these capacitors is computed. The results of this process are presented.

   INTEGRATED CONDITION MONITORING AND DIAGNOSIS OF ELECTRICAL MACHINES USING MINIMUM CONFIGURATION ARTIFICIAL INTELLIGENCE
By F. Filippetti; C. Tassoni; G. Franceschini; P. Vas [View]
[Download]

Abstract: The traditional approach to electrical machine condition monitoring based on deterministic models leads to complex systems hard to maintain and manage. Artificial Intelligence (AI) techniques seem to offer a simpler solution which avoids the use of deterministic models. An effective and simple system acceptable in industrial environments requires a "minimum configuration intelligence". This paper deals with the application of AI techniques for diagnosis purposes. In particular, new results related to the application of fault diagnosis of electrical machines using Neural Networks with a minimum number of neurons and Fuzzy-Neuro (ANFIS) techniques are presented

   DETERMINISTIC APPROACH AND NEURAL NETWORK APPROACH FOR STATOR SHORT CIRCUITS DIAGNOSIS IN INDUCTION MACHINES
By F. Filippetti; G. Franceschini; C. Tassoni; G. Gentile; S.Meo; A. Ometto; N. Rotondale [View]
[Download]

Abstract: In this paper short circuits in stator windings of induction machines are considered. Two diagnostic approaches are analysed. They are respectively based on different deterministic models of the faulted machine and on neural network techniques to represent the system behaviour. The deterministic faulted models have an increasing simplification level and consequently a decreasing effectiveness. About the neural network approach the crucial points are the input/output retrieving, the choice of input and output variables and the learning phase. The relevant performances obtained by means of the two approaches are compared and discussed.

   FAULT DETECTION AND DIAGNOSIS IN THE ELECTRIC DRIVES
By R. Szczesny; H. Piquet; P. Kurzynski [View]
[Download]

Abstract: In this paper is presented an approach for fault diagnosis in an induction motor drive. In order to recognize the default, both time-domain and frequency-domain are considered. The first method is mainly applied when the converters are suspected to have abnormal behavior: starting from the measured waveforms of the voltage and the currents at the converters inputs and output, it is able to point out which switch of the converter is out of order. The second method uses harmonics values of a set of signals of the system and can be applied more generally. The choice of the method depends on the number of available sensors. Implementation is presented and discussed for both methods and has been.achieved by mean of SMECI - an expert system shell.

   MODELLING AND SIMULATION OF SEMICONDUCTOR SYSTEMS FOR DIAGNOSTIC PURPOSES
By Zygfryd Glowacz [View]
[Download]

Abstract: Stiff differential equations of mathematical models of semiconductor systems are solved using SESL2 and FORTRAN languages. The electromechanical rolling mill system with 2H-type converter is considered as characteristic semiconductor system. The influences of pulses missing in thyristor control, shorting of thyristor, loss of thyristor blocking ability on characteristic quantities of electromechanical rolling mill system, are determined. The calculation results are presented in a form of diagrams of electromechanical system responses. A comparison of calculation and measurement results is performed.

EPE 1997 – 49: Dialogue Session DS9f: DIAGNOSTICS
You are here: EPE Documents > 01 - EPE & EPE ECCE Conference Proceedings > EPE 1997 - Conference > EPE 1997 – 49: Dialogue Session DS9f: DIAGNOSTICS
	[return to parent folder]

	USE OF THE OUTPUT VOLTAGE OF A SWITCHMODE POWER SUPPLY TO PREDICT FAILURE OF FILTERING ELECTROLYTIC CAPACITORS By P. Venet; A. Lahyani; P.J. Viverge; G. Grellet	[View] [Download]
Abstract: Filtering electrolytic capacitors are responsible of most breakdown cases in static converters. The best indicator of fault of these capacitors is the equivalent series resistance (ESR) which increases versus temperature and ageing time according to a well known law. Since the output voltage ripple of the converter is almost proportional to ESR, the failure of the capacitors is predicted by monitoring online this voltage during operation of the converter. So as to avoid errors due to sudden load variation, the output voltage ripple is firstly filtered at the switching frequency of the converter, then processed with other converter parameters, thereby, the lifetime of these capacitors is computed. The results of this process are presented.

	INTEGRATED CONDITION MONITORING AND DIAGNOSIS OF ELECTRICAL MACHINES USING MINIMUM CONFIGURATION ARTIFICIAL INTELLIGENCE By F. Filippetti; C. Tassoni; G. Franceschini; P. Vas	[View] [Download]
Abstract: The traditional approach to electrical machine condition monitoring based on deterministic models leads to complex systems hard to maintain and manage. Artificial Intelligence (AI) techniques seem to offer a simpler solution which avoids the use of deterministic models. An effective and simple system acceptable in industrial environments requires a "minimum configuration intelligence". This paper deals with the application of AI techniques for diagnosis purposes. In particular, new results related to the application of fault diagnosis of electrical machines using Neural Networks with a minimum number of neurons and Fuzzy-Neuro (ANFIS) techniques are presented

	DETERMINISTIC APPROACH AND NEURAL NETWORK APPROACH FOR STATOR SHORT CIRCUITS DIAGNOSIS IN INDUCTION MACHINES By F. Filippetti; G. Franceschini; C. Tassoni; G. Gentile; S.Meo; A. Ometto; N. Rotondale	[View] [Download]
Abstract: In this paper short circuits in stator windings of induction machines are considered. Two diagnostic approaches are analysed. They are respectively based on different deterministic models of the faulted machine and on neural network techniques to represent the system behaviour. The deterministic faulted models have an increasing simplification level and consequently a decreasing effectiveness. About the neural network approach the crucial points are the input/output retrieving, the choice of input and output variables and the learning phase. The relevant performances obtained by means of the two approaches are compared and discussed.

	FAULT DETECTION AND DIAGNOSIS IN THE ELECTRIC DRIVES By R. Szczesny; H. Piquet; P. Kurzynski	[View] [Download]
Abstract: In this paper is presented an approach for fault diagnosis in an induction motor drive. In order to recognize the default, both time-domain and frequency-domain are considered. The first method is mainly applied when the converters are suspected to have abnormal behavior: starting from the measured waveforms of the voltage and the currents at the converters inputs and output, it is able to point out which switch of the converter is out of order. The second method uses harmonics values of a set of signals of the system and can be applied more generally. The choice of the method depends on the number of available sensors. Implementation is presented and discussed for both methods and has been.achieved by mean of SMECI - an expert system shell.

	MODELLING AND SIMULATION OF SEMICONDUCTOR SYSTEMS FOR DIAGNOSTIC PURPOSES By Zygfryd Glowacz	[View] [Download]
Abstract: Stiff differential equations of mathematical models of semiconductor systems are solved using SESL2 and FORTRAN languages. The electromechanical rolling mill system with 2H-type converter is considered as characteristic semiconductor system. The influences of pulses missing in thyristor control, shorting of thyristor, loss of thyristor blocking ability on characteristic quantities of electromechanical rolling mill system, are determined. The calculation results are presented in a form of diagrams of electromechanical system responses. A comparison of calculation and measurement results is performed.