Integration of knowledge-based, qualitative and numeric tools for real time dynamic systems supervision

Abstract

The proposal presented in this thesis is to provide designers of knowledge based supervisory systems of dynamic systems with a framework to facilitate their tasks avoiding interface problems among tools, data flow and management.<br/>The approach is thought to be useful to both control and process engineers in assisting their tasks. The use of AI technologies to diagnose and perform control loops and, of course, assist process supervisory tasks such as fault detection and diagnose, are in the scope of this work. Special effort has been put in integration<br/>of tools for assisting expert supervisory systems design. With this aim the experience of Computer Aided Control Systems Design (CACSD) frameworks<br/>have been analysed and used to design a Computer Aided Supervisory Systems (CASSD) framework. In this sense, some basic facilities are required to be available in this proposed framework:<br/>·

ion Tools, for signal processing,<br/>representation and analysis to obtain significative information.<br/>· To deal with process variables, measures or numerical estimations, and expert observations, with uncertainty and imprecision.<br/>· Expert knowledge representation at different levels by using a rule-based system or simple qualitative relations.<br/>· Modularity and encapsulation of data and knowledge would be useful for structuring information.<br/>· Graphical user interface to manage all those facilities in the same environment as actual CACSD packages.<br/>Several tools from the AI domain have been added as Simulink ToolBoxes to deal with abstracted information, qualitative relationship and rule-based ES. Simple and intuitive qualitative relationship can be implemented by means of ablock-based qualitative representation language called ALCMEN. An ES shell, called CEES, has also been embedded into MATLAB/Simulink as a block to<br/>allow modularisation and partition of large expert KBs. Finally, the numeric to qualitative interfaces is performed by a set of algorithms, called abstraction tools, encapsulated also in Simulink blocks. The functionality of the whole<br/>framework is able due to the use of object oriented approach in the development and implementation of those tools.<br/>In this thesis an attempt is undertaken to make steps towards integration of tools for expert supervision, including once for qualitative and symbolic data representation and management and symbolic knowledge processing. The main research objectives of this work include the following points :<br/>1. Incorporation of object-variables into classical numerical data processing system. The aim is to allow structural qualitative and symbolic knowledge representation. Complex information is encapsulated in a single source/sink structure, called object-variable, providing methods for knowledge access and processing.<br/>2. Implementation of selected particular tools for qualitative and symbolic knowledge representation and interfacing. Higher abstract level information processing based on the introduced object-variables.<br/>3. Embedding an object oriented rule-based expert system into a classical CACSD framework in order to provide high level knowledge processing facilities based on the domain of expert knowledge, heuristics, and logic.<br/>The object approach forces engineers to structure knowledge becoming highly locatable, modular and encapsulated. This features are very important to getexpert supervisory system design closer to process. The objective is to approach design tools to process engineers avoiding extra-time in learning application functionality and interfacing process variables and design tools. Thus, objects are used in the process variables descriptions as sources of information, encapsulating tools to provide significant (qualitative or numerical) information. Object oriented features will permit to divide large KBs into smaller ones to deal with complex systems adopting distributed solutions. Consequently, ES becomes more specialised, maintainable, and easier to validate.