Safety, supervision and diagnosis of industrial plants

Sylvie CHARBONNIER ( ).

Objectives

The objective of this class is to introduce the concept of fault detection and fault diagnosis for complex systems and to present different classes of methods which have proven their performances in practical applications.

Main topics

Lesson Topic
1
Introduction to supervision
Tasks of supervision, terminology.
2
Model-based fault detection
Parity equations, observers, on line estimation of model parameters.
3
Signal-based fault detection
Features extraction using time, frequency and time-frequency transformation, pattern comparison. Temporal change detection.
4
Data-driven fault detection methods
Fault diagnosis with pattern recognition, fault diagnosis with principal component analysis.
SUPERVISION LABS
Lab 1
Fault detection in a two-tanks system using a bank of observers
Lab 2
On-line detection of deep sleep using EEG spectral power
Lab 3
Diagnosis of a mineral treatment unit using pattern recognition
Lab 4
Sensor fault detection in an air quality monitoring network using principal component analysis

References

  • S. Gentil (Ed.), "Supervision des procédés complexes", HERMES Systèmes automatisés, 2007.
  • Isermann, "Fault diagnosis systems", Springer, 2006.
  • Blanke, Kinnaert, Lunze, Staroswiecki, "Diagnosis and fault tolerant control", Springer, 2003.

Grading policy

Labs: 100 %

Handouts

Restricted access area