On Dependability in Distributed Computerized Manufacturing Systems

Doctoral thesis, 1993

The concepts of dependability, fault tolerance, reliability and availability concerning computer systems had already emerged by the 1950's. These concepts were initially used in the area of computer hardware, but were then applied to the area of software and later to applications. This thesis is focused on the application of different aspects of fault tolerance to distributed computerized manufacturing systems. Its aim is to contribute to the almost new application field of fault tolerant manufacturing. The work contained in this thesis has followed four main directions. Work related to the first direction focus on the definition of a hierarchical tool called GRS (General Recursive System) that is used to model fault tolerance in hierarchical computerized manufacturing systems. Work related to the second direction has to do with system reconfiguration and its use to improve fault tolerance. In the work on this area, an adequate taxonomy is defined that describes in detail various aspects of fault tolerance in conjunction with manufacturing system reconfiguration. This taxonomy is applied in a case study of a flexible assembly line. Work related to the third direction focus on the accessibility of information and its importance in obtaining a fault tolerant system. In this work, it is shown how the use of cache memories in a manufacturing system can enhance fault tolerance in the case of data network partitionings. Another aspect discussed is the possible use of the material network as a source of data transportation. Both aspects are applied to the case study example that was used in the work performed in the area of the second direction. Work related to the fourth direction, which has only recently been initiated, relates to failure semantics in distributed computerized manufacturing systems. It is demonstrated in the work carried out in this area that a concise definition of the failure semantics for a manufacturing system is essential in order to obtain correct fault tolerance.