EPE Association: EPE 1999 - Topic 10e: Diagnostics

EPE 1999 - Topic 10e: Diagnostics
You are here: EPE Documents > 01 - EPE & EPE ECCE Conference Proceedings > EPE 1999 - Conference > EPE 1999 - Topic 10: SYSTEMS ENGINEERING > EPE 1999 - Topic 10e: Diagnostics

   [return to parent folder]

   Fault Detection Evaluation of Microcontroller Dyad Control System by Fault Injection Method
By Ž. Hocenski; G. Martinovic [View]
[Download]

Abstract: The fault detection methods are very important in the fault-tolerant system design. The redundancy is the basis for the fault tolerance, but it is used in the fault detection also. One of the basic fault detection methods is the duplication with comparing. The duplication could be done on several levels in hardware and software. The control system based on microcontrollers in dyad are presented in this paper. The microcontrollers in dyad are used for increased reliability and fault tolerance. They are loosely coupled with task synchronization at checkpoints. The results of operation are exchanged and compared at a checkpoint. Depending on two types of software the behavior of such duplicated system could be done for the increased availability or for the increased safety. By the safe application both units must reach an agreement at each checkpoint, otherwise the system outputs will be disabled. In a highly available application a valid unit continues to work after disagreement at a checkpoint. Therefore both units must have the self-checking capabilities. The evaluation of used fault detection methods and fault tolerance is done by experiments using the fault injection method. The fault injection system is based on a personal computer, which controls the experiment and collects the results. The bus signals activity is used in calculation of the probability of the fault occurrence. The fault recovery coverage is evaluated based on the registered number of faults and the probability of the fault occurrence.

   Wavelet and Neural Network Structure for Analyzing and Classifyin...
By M. Castilla; D. Borrás; N. Moreno; J.C. Montaño [View]
[Download]

Abstract: Discrete wavelet transform (DWT) and artificial neural networks (ANN) are used for detecting, compressing and classifying power quality disturbances. The method begins by decomposing an input signal into its details, and its most-smoothed signal. The detail signals contain wavelet transform coefficients (WTC). Thresholding of WTC permits selection of those corresponding to disturbance events. To recover the input signal, reconstruction is performed using the most-smoothed signal, along with the saved WTC of the detail signals. Data are stored with a high compression ratio, while the error between the input and the reconstructed signals is minimized. Eight types of actual power line disturbances are classified using an ANN structure.

EPE 1999 - Topic 10e: Diagnostics
You are here: EPE Documents > 01 - EPE & EPE ECCE Conference Proceedings > EPE 1999 - Conference > EPE 1999 - Topic 10: SYSTEMS ENGINEERING > EPE 1999 - Topic 10e: Diagnostics
	[return to parent folder]

	Fault Detection Evaluation of Microcontroller Dyad Control System by Fault Injection Method By Ž. Hocenski; G. Martinovic	[View] [Download]
Abstract: The fault detection methods are very important in the fault-tolerant system design. The redundancy is the basis for the fault tolerance, but it is used in the fault detection also. One of the basic fault detection methods is the duplication with comparing. The duplication could be done on several levels in hardware and software. The control system based on microcontrollers in dyad are presented in this paper. The microcontrollers in dyad are used for increased reliability and fault tolerance. They are loosely coupled with task synchronization at checkpoints. The results of operation are exchanged and compared at a checkpoint. Depending on two types of software the behavior of such duplicated system could be done for the increased availability or for the increased safety. By the safe application both units must reach an agreement at each checkpoint, otherwise the system outputs will be disabled. In a highly available application a valid unit continues to work after disagreement at a checkpoint. Therefore both units must have the self-checking capabilities. The evaluation of used fault detection methods and fault tolerance is done by experiments using the fault injection method. The fault injection system is based on a personal computer, which controls the experiment and collects the results. The bus signals activity is used in calculation of the probability of the fault occurrence. The fault recovery coverage is evaluated based on the registered number of faults and the probability of the fault occurrence.

	Wavelet and Neural Network Structure for Analyzing and Classifyin... By M. Castilla; D. Borrás; N. Moreno; J.C. Montaño	[View] [Download]
Abstract: Discrete wavelet transform (DWT) and artificial neural networks (ANN) are used for detecting, compressing and classifying power quality disturbances. The method begins by decomposing an input signal into its details, and its most-smoothed signal. The detail signals contain wavelet transform coefficients (WTC). Thresholding of WTC permits selection of those corresponding to disturbance events. To recover the input signal, reconstruction is performed using the most-smoothed signal, along with the saved WTC of the detail signals. Data are stored with a high compression ratio, while the error between the input and the reconstructed signals is minimized. Eight types of actual power line disturbances are classified using an ANN structure.