392485 Functional Modelling and Reasoning about Safety in Complex Process Plant

Wednesday, November 19, 2014: 8:55 AM
404 - 405 (Hilton Atlanta)
Morten Lind, Department of Electrical Engineering, Technical University of Denmark (DTU), Lyngby, Denmark

Decision support systems can play a significant role in in the management of safety critical systems by providing facilities for data acquisition, interpretation and information sharing. Knowledge based systems has been proposed by several researchers as tools for implementation of intelligent decision support. Rule bases and qualitative models encoding expert knowledge about the process are here used by inference engines to give advice to decision makers in the analysis of complex problems. Within process safety, such intelligent systems are considered for computer assisted Hazop and for online operator support in fault diagnosis and planning of remedial actions.

Applications of knowledge based technology show that their success depends critically on the nature of the knowledge available and of the problem to be solved. Thus early applications of rule bases showed that the encoding of expert knowledge also should include concepts and relations which could not be formulated in rules but should be expressed in models. The choice of ontologies to represent concepts and relations which are expressed in the models are therefore important in present developments of knowledge based systems, which often include both rules and models.

Modeling and qualitative reasoning about safety problems are typically based on cause-effect relations. For example cause-consequence trees which can be expressed by rules, are explicit representations of the relations between process faults and their causes and consequences.  In model based systems for reasoning about process safety, faults have no explicit representation but they are derived from a model of process elements and a rule-base containing generic knowledge about cause-effect relations between the elements.  

The paper will present a modeling and reasoning technique based on means-end concepts called Multilevel Flow modelling (MFM). MFM is a technique for functional modeling. Its main focus on relations between means and ends or goals (rather than functions) provides a unique framework for process representation on multiple levels of function and for modeling levels of defense. MFM concepts support qualitative modelling of processes with interacting material, energy and information flows and has been used to model nuclear power plants, power systems and chemical engineering plants. MFM has been applied for cause-consequence analysis in Hazop and for online diagnosis.

The paper will introduce the basic concepts of MFM and present modelling examples from nuclear power and from a recent Hazop related study of an oil and gas separation plant. A short introduction to the MFMStudio, a model based reasoning tool, will also be presented demonstrating the capability of modeling technique and the implemented rule bases for cause-consequence reasoning.

Qualitative modelling and reasoning as implemented with MFM can often with advantage be combined with other types of models in order to increase coverage of the analysis or for validation purposes. In the Hazop study mentioned above, it has been investigated how the reasoning results from an MFM model could be validated by comparing it with simulations using a quantitative model (differential equations). The two models represent different types of knowledge (goals and functions versus physical interdependencies) and support different inferences. Preliminary results and some implications of this validation study will be discussed.

Extended Abstract: File Not Uploaded