
Analog Circuit Simulation and Troubleshooting with FLAMES?
F. MOHAMEDy, M. MARZOUKI A. BIASSIZO, F. NOVAK
TIMA Laboratory, 46 avenue F?elix Viallet, Jozef Stefan Institute, Jamova 39,
38031 GRENOBLE Cedex, FRANCE 61111 Ljubljana,Slovenia
Abstract
A new approach for analog circuit simulation and troubleshooting, based on the fuzzy logic paradigm, is presented in this paper. This approach allows to deal with soft faults, single or multiple ones, and with both impreciseness and uncertainty of information. It has been implemented in a system named FLAMES (Fuzzy Logic ATMS and Modelbased Expert System), which details are provided, together with different experimental results, for both simulation and troubleshooting process.
1 Introduction
FLAMES [1], short for Fuzzy Logic ATMS and Modelbased Expert System, is a software developed to deal with analog system troubleshooting and specially with analog circuits with their known difficult problems.
The research in this area has been driven towards AI [2] approaches which may be able to get it out of this bottleneck. Modelbased reasoning, qualitative reasoning, and other approaches were the basis of many systems which have tried to deal with analog troubleshooting [1].
In this paper, we present a new approach mainly based on fuzzy logic and modelbased reasoning. A fuzzy ATMS (Assumption Truth Maintenance System) is the core of this approach ; it propagates fuzzy intervals, which we show that they are more suitable for fault detection and fuzzy assumptions. It has some important properties : first, it deals with multiple faults; second, it is not necessary for the starting points to be inputs or outputs of the circuit; third, a best test strategy finding process, based on fuzzy entropy and fuzzy decisionmaking which does not need a heavy calculus as in other used approaches, is developed. Furthermore, FLAMES could be used for other purposes than troubleshooting, such as simulation.
?This work was partly supported by the FrenchSlovenian
PROTEUS Project #95010
yPartly supportedby a grant from HIAST, the Syrian Higher
Institute for Applied Sciences and Technology
Section 2 presents fuzzy logic which is the mathematical basis of our approach. Section 3 discusses the main ideas among the reasons of our research directions, and the general diagram of our system (FLAMES) is presented in section 4. Section 5 details the fuzzy ATMS, while sections 6 and 7 describe two other components of FLAMES, respectively dealing with learning from experience and best test strategy finding. The implementation is detailed in section 8, and experimental results are provided, both in case of troubleshooting and in case of simulation. Finally, conclusion and future work directions are given in section 9.
2 Fuzzy logic
The principal objective of fuzzy logic is to cope with the inaccuracy and the uncertainty of information.
A fuzzy set S is defined on a domain T by a function ?S:T >[0,1], such that ?S(t) is the membership degree of t 2 T to the set S. A fuzzy interval is a convex fuzzy set. In practice, it will be defined by a 4tuple [m1,m2,ff; ] [3], where [m1,m2] is its core. This representation allows a crisp number, a crisp interval, a fuzzy number, and a fuzzy interval to be uniformly described and it also offers an efficient arithmetic treatment.
3 Towards Fuzzy Logic for the Analog
Diagnosis
Some systems have been developed for analog circuit
fault diagnosis. The most representative ones are
briefly described in the following.
DEDALE [4] is based on orderofmagnitude reasoning. Its main weakness appears when dealing with components which operate at the limit of their designed behavior because it assumes that defects lead to significant changes in behavior of the circuit which is a hard assumption. In DIANA [5] imprecision is processed by means of numerical (crisp) intervals. The management of intervals is done by an ATMS extension. FIS [6], uses qualitative causal models to describe the unit under test. It is suitable to deal with analog systems but we believe that it could suffer from