Estudia

1. (1 h) Presentation of the course.

2. (1 h) Introduction to fault detection and diagnosis. Objective. Concepts, terminology: fault. Tasks: detection, diagnosis, fault tolerance. Characteristics: false alarms, missed alarms, detectability, diagnosability.

3. (1 h) Techniques for fault detection and diagnosis: model-based, statistical, artificial intelligence.

4. (1 h) Data driven methods for fault detection and diagnosis. Data mining point of view.

5. (2 h) Univariate Statistical Process Control

6. (4 h) Multivariate Statistical Process Control (PCA/PLS, DPCA, MPCA)

7. (2 h) Artificial Intelligence based methods for Fault Detection and diagnosis (Case based reasoing, Artificial Neural Networks, Sequences and similarity, etc)

8. (1 h) Introduction to model-based fault detection and diagnosis techniques. Concepts: system, Systems Theory, model, experiment.

9. (1 h) Model types: physical, mental, symbolic, mathematical, no mathematical. Models obtaining: modelling and identification.

10. (1 h) Model types: static and dynamic, time variant and time invariant, lumped parameters and distributed parameters, deterministic and probabilistic, continuous and discrete, linear and non linear. Linearisation.

11. (2 h) Physical redundancy. Analytical redundancy. Classification of fault detection and diagnosis techniques. Model-based fault detection and diagnosis: FDI and DX scientific communities. Residuals. Techniques for residual generation: State Observers, Parity Equations, Parameter Estimation. Structural Analysis. Causal Graphs.

12. (2 h) Residual evaluation. FDI Diagnosis. Fault signature. Fault isolation: directed and structured residuals. An example of a sensor fault. An example of a process fault.

13. (1 h) DX diagnosis: consistency based diagnosis. BRIDGE: integration of FDI and DX techniques. Comparison FDI-DX.

14. (1 h) Fault detection in uncertain systems. Uncertainty: precision, accuracy. Imprecise models. Quantitative, semiqualitative and qualitative models. Models of systems with parametric uncertainty. Interval models.

15. (1 h) Simulation of uncertain models: generation of envelopes. Properties of envelopes. Aplication of envelopes to fault detection. Quantitative, semiqualitative and qualitative simulators.

16. (1 h) Intervals. Simulation of interval models. Generation of the exact envelope. Modal intervals. Generation of bounded-error envelopes. Sliding time windows. f* algorithm.

17. (1 h) Fault detection using envelopes: SQualTrack.

• Russell, Evan (cop. 2000 ). Data-driven methods for fault detection and diagnosis in chemical processes . London [etc.]: Springer. Catàleg
• Eriksson, L. (2006 ). Multi- and megavariate data analysis (2nd rev. and enl. ed.). Umea: Umetrics AB. Catàleg
• Adamo, Jean-Marc (2000 ). Data mining for association rules and sequential patterns : sequential and parallel algorithms . New York: Springer. Catàleg
• Witten, I. H. (cop. 2005 ). Data mining : practical machine learning tools and techniques (2nd ed.). Amsterdam [etc.]: Morgan Kaufman. Catàleg
• Cellier, François E. (1991 ). Continuous system modeling . New York [etc.]: Springer-Verlag. Catàleg
• Cellier, François E. (1991). Continuous system modeling. . Recuperat , a http://people.inf.ethz.ch/fcellier/Pubs/BG/springer.html Catàleg
• Gertler, Janos J. (c1998 ). Fault detection and diagnosis in engineering systems . New York [etc.]: Marcel Dekker. Catàleg
• Chen, Jie (1999 ). Robust model-based fault diagnosis for dynamic systems . Boston [etc.]: Kluwer Academic Publishers. Catàleg
• Patton, Ron J. Frank, Paul M. Clark, Robert (cop. 2000 ). Issues of fault diagnosis for dynamic systems . London [etc.]: Springer. Catàleg
• Blanke, Mogens Schröder, Jochen (cop. 2003 ). Diagnosis and fault-tolerant control . Berlin [etc.]: Springer. Catàleg
• Ding, Steven X (2008 ). Model-based fault diagnosis techniques : design schemes, algorithms, and tools . New York: Springer. Catàleg

Activitats d'avaluació:

If mark_work>=4 and mark_exam>=4 then mark_subject=0,1*mark_P1 + 0,05*mark_P2 + 0,05*mark_P3 + 0,4*mark_work + 0,4*mark_exam.

Criteris específics de la nota «No Presentat»:
La nota de l'assignatura serà No Presentat en qualsevol d'aquests casos:
- Hi ha faltes d'assistència a les pràctiques.
- No s'entrega el treball pràctic com a mínim 10 dies abans de l'examen.
- No s'entrega l'examen.