Тестирование систем с нечеткими дискретными компонентами.

Abstract

Modern complex systems are build based on heterogeneous components with various interrela tionships, fuzziness, and uncertainty of the laws of functioning of the components and the system. An important class of such systems comprises hybrid intelligent systems, where the components are represented by analytical models of fuzzy objects, artifi cial neural networks, expert systems, etc. The article considers fuzzy discrete devices being, for example, part of hybrid systems. Fuzzy linear automata (FLA) introduced in the article are used as a mathematical model of such components. The problem of test synthesis for FLA used to detect faults in them is discussed. Normal single-stuck faults are permissible faults in FLA. The faults originating from the replacement of some elements of the FLA characteristic matrices with others (from a given set of alternative ones) are also permissible. Test synthesis methods for FLA belonging to the class of m-deterministic and synchronized automata, as well as arbitrary linear automata have been developed. The fi rst two methods are based on reducing the considered problem of solving systems of linear algebraic equations. It should be noted that there is a well-developed mathematical apparatus applying a few eff ective methods for searching for such solutions. The tests synthesized by these methods for m-deterministic and synchronized FLA are sufficiently short and do not exceed the memory depth of the corresponding automata. It is shown that the conditions for an FLA referring to the two fi rst classes mentioned above are not too strict. It is noted that the known methods of test synthesis for linear automata require compliance with much more stringent requirements. The synthesis method for arbitrary FLA also builds short tests