Browsing by Author "Rebolledo, Mario"

Now showing 1 - 1 of 1

Open Access
Situation based process monitoring in complex systems considering vagueness and uncertainty
(2004) Rebolledo, Mario; Göhner, Peter (Prof. Dr.-Ing. Dr. h. c.)
History has demonstrated during the 20th century that industrial development carries hazards that should not be ignored because they endanger humans, the environment and production facilities. For this reason, continuous development of new production technologies should be accompanied by a comparable development in industrial safety technologies. Safety-critical applications in complex processes are usually based on a precise monitoring of operation conditions, according to a “correct” process operation. The problem is determining if a behavior or an operation condition is “correct”. For this, models are generally used, which are able of reproducing “safe” or “appropriate” process behaviors. The difficulty of precise modeling grows continuously, because of the increasing complexity of the supervised processes. Rigorous deterministic modeling is limited to simple processes, while approximate models based on statistics or Artificial Intelligence techniques, for example, must be restricted to modeling single variables or small subsystems to be manageable and deliver useful information. A monitoring technique usually employed for complex processes relies on abstraction of the process behavior in qualitative models by using symbolic value ranges to represent required information. However, also the applicability of qualitative modeling techniques is eventually restricted by the resulting model size. In this research work, a new process monitoring approach, based on qualitative models, efficiently depicts valuable vague and uncertain information that is currently discarded during the modeling. The proposed method expands the ability of Situation-based Qualitative Modeling and Analysis (SQMA) to monitor complex processes by integrating elements of the Rough Set Theory and Stochastic Qualitative Automata. The resulting models are considerably more precise than other similar-sized qualitative models. At the same time, the new method develops more compact and precise qualitative models than traditional qualitative models of the same precision.