ePrints@IIScePrints@IISc Home | About | Browse | Latest Additions | Advanced Search | Contact | Help

Fault diagnosis using signed digraphs: a neural network approach

Sharma, SM and Viswanadham, N (1991) Fault diagnosis using signed digraphs: a neural network approach. In: Proceedings of the SPIE - The International Society for Optical Engineering, 2-5 Jan. 1991, Bangalore, India, pp. 675-685.

Full text not available from this repository.

Abstract

The problem of locating the origin of failure in a dynamical system using signed digraphs is of exponential complexity. It is therefore attractive to use the parallel architecture of a neural network to solve the problem. The authors formulate this problem as a constrained quadratic optimization problem. This formulation enables them to synthesize a neural network that solves this constrained optimization problem and thus the fault diagnosis problem. This approach to fault diagnosis is attractive because the neural network employed takes advantage of distributed information processing and its inherent ability to perform parallel computation. Also, the hardware employed for fault diagnosis is robust, as the malfunctioning of a single component does not lead to serious breakdown. The structure of the network used is such that it can be easily implemented on VLSI chips. The space complexity of the network is of the order O(m+n) where m and n are the number of nodes and edges respectively. The applicability of this method is demonstrated by means of an example.

Item Type: Conference Paper
Additional Information: Copyright of this article belongs to SPIE - The International Society for Optical Engineering.
Keywords: computational complexity;computer aided analysis;directed graphs;failure analysis;neural nets;parallel architectures; quadratic programming
Department/Centre: Division of Electrical Sciences > Computer Science & Automation (Formerly, School of Automation)
Date Deposited: 18 Jan 2008
Last Modified: 12 Jan 2012 09:18
URI: http://eprints.iisc.ernet.in/id/eprint/11192

Actions (login required)

View Item View Item