DAI-DEPUR An integrated Supervisory Multi-level Architecture.

Author

Miquel Sànchez-Marrè miquel@lsi.upc.es

Departament de Llenguatges i Sistemes Informàtics
Universitat Politècnica de Catalunya
Edifici C5-Campus Nord
C/ Jordi Girona 1-3
08034 Barcelona, Catalonia, Spain

Abstract

In this thesis, it is presented the research and work developed in the design and implementation of an integrated multi-level architecture for wastewater treatment plants (WWTPs) supervision in real-time. The research has coped with a double open problems in two different areas: the insufficiency of classical Chemical Engineering control methods applied to WWTPs, and on the other hand, some pitfalls of Knowledge-Based Systems, specially when faced against real-world problems. The management, control and supervision of a WWTP is a very complex and dangerous task, due to the features of a WWTP and to the catastrophic consequences that can be achieved by an incorrect WWTP operation. Usual used techniques –numerical control algorithms– are not able to control the WWTP if it is not operating in normal conditions. They need the integration of other techniques that allow to include the expert knowledge provided by the WWTP's operators and the literature, and the experiential knowledge acquired in the past operation of the WWTP under supervision. Knowledge-Based Systems (KBS) have some pitfalls, specially when faced against complex real-world domains: their scope is limited to the forecasted situations in the domain, i.e. brittleness; most KBS do not learn from their experiences; the knowledge acquisition problem; low reusability of knowledge bases, and the increasing complexity of monolithic problem solving systems The architecture, called DAI-DEPUR, is the result of the integration of several Artificial Intelligence techniques with some Control Engineering methods, and with some Chemical Engineering techniques: numerical control methods – a predictive control algorithm– , Chemical Engineering models, rule-based reasoning, case-based reasoning, semi automated knowledge acquisition, learning, on-line and off-line data acquisition, etc . The global issue of supervision is carried out in a distributed way by means of several tasks: system evaluation, local diagnosis of subsystems, adaptation, global diagnosis, supervision, operator's validation and actuation. The expert knowledge is distributed among several knowledge bases that cooperate for the global supervisory task. The architecture is multi-level , and it has been structured in this way, as a result of the study of the different kinds of knowledge and tasks involved in the domain. This feature provides it with a certain independence among the different levels: the data level, the knowledge/expertise level, the situations level and the plans level. DAI-DEPUR has been implemented by means of some tools such as G2 – a real-time expert systems shell– , LINNEO +– a semi-automated unsupervised knowledge acquisition tool– , GAR – an inference rule automated generator– , and the programming language Lisp for the implementation of the case-based reasoner. The evaluation of the system has given good results and it has been carried out in two stages. First, the three main components of the architecture: the numerical control knowledge, the expert knowledge and the experiential knowledge. Secondly, a global validation of DAI-DEPUR, also containing three steps has followed: WWTP operation simulations validated by the experts, validation in a pilot scale WWTP constructed to that end, and the next evaluation in a real WWTP by means of an agreement with the "Junta de Sanejament de la Generalitat de Catalunya". Finally, some features of DAI-DEPUR execution are showed, and a few examples of application are detailed, in order to outline the global supervisory process where interact the several techniques implemented, the WWTP and the WWTP's operator through several interfaces.

Keywords: Integrated Architectures, Multi-level Architectures, Distributed Architectures, Rule-based reasoning, Case-based reasoning, Knowledge acquisition, Learning, Real-time Supervision and Control, Wastewater treatment, Biotechnology, Chemical Engineering, Environmental Engineering.

Citation:

M. Sànchez-Marrè, "DAI-DEPUR: An Integrated Supervisory Multi-Level Architecture for Wastewater Treatment Plants" . PhD thesis. Departament de Llenguatges i Sistemes Informàtics. Universitat Politècnica de Cantalunya, 1996.

World Wide Web

You may want to view a WWW version of the PhD thesis: DAI-DEPUR An integrated Supervisory Multi-level Architecture.

Postscript

You may want a postcript version of the PhD thesis:
  1. Cover
  2. Index
  3. Introduction
  4. The State of the Art
  5. DAI-DEPUR: an Integrated Supervisory Multi-level Architecture
  6. The Data Level
  7. The Knowledge/Expertise Level
  8. The Situations Level
  9. The Plans Level
  10. Experimental Evaluation and Validation
  11. Application
  12. Conclusions and Future Work
  13. Glossary
  14. Tools
  15. Bibliography


Back to the Artifical Intelligence Section page


Last modified on .

Page mantainer: webia@lsi.upc.es