Diagnosis and Supervision

Aim

To give both a theoretical and practical basis for how to design systems that automatically detect and isolate faulty components in technical processes. After the course is finished, the student shall:

• know why diagnosis systems are used in different industrial application areas.
• know how to analyze which faults in a complex process that need to be supervised to achieve the overall goals.
• from a case description be able to structure the problem and develop principle and architechture for a complete implementation of a diagnosis system.
• given a formal model description be able to choose suitable mathemematical methods to solve the problem.
• know advantages and disadvantages of the different methods that are included in the course.
• be able to apply mathematical tools and methods from a variety of previous courses to solve diagnosis problems.
• be able to value and verify functionality and performance of a complete diagnosis system.
• have a broad theoretical insight in the subject, deep enough to be able to understand and utilize new research results developed by the research community.

Content

• Introduction: history and overview, practical application examples.
• Principles for model based diagnosis: mathematical modelling of faults, detection and isolation of faults by means of models, consistency relations, analytical redundancy, decisions with structured hypothesis tests.
• Control theory methods: linear and non-linear residual generation, observers and Kalman filters for diagnosis, residual evaluation, adaptive thresholding, statistical methods.
• Logic based AI methods: basic principles, fault isolation algorithms.
• Probability based diagnosis and Bayesian networks.
• Other: fault trees and FMEA, statistical methods/change detection.