A Fast Algorithm for Testability Analysis of Large Linear Time-Invariant Networks

Testability analysis is aimed at evaluating the degree of solvability of both the fault diagnosis and the parameter identification problems. As far as analog linear time-invariant circuits are concerned, various approaches, predicated on either fully numerical or symbolic techniques, have been proposed in the past for such purpose. However, the former are severely plagued by the effects of roundoff errors, whereas the latter soon become unacceptably time-consuming as the circuit size grows large. In the present paper, a novel algorithm is described, which completely avoids resorting to symbolic techniques, thereby overcoming the mentioned related drawbacks and, at the same time, it drastically reduces the computational issues posed by formerly proposed fully numerical methods. Such an algorithm provides information as to how many and which parameters can be diagnosed or identified and has been straightforwardly translated into a software for fully automated testability analysis, also briefly described herein. The aforementioned noteworthy features of the algorithm on which it is based render the software in question a successful tool for testability analysis of very large networks, as opposed to formerly proposed computer programs, which generally fail such a task. Examples of application of both the algorithm and its software counterpart are given.