Course and Exam Registration
Please use the course and exam registration form to enrol in the course and register for the final examination.
If you have any problems accessing the form, please contact Prof. Silvio Simani at silvio.simani@unife.it.
Prerequisites
Participants are expected to have basic knowledge of automatic control systems and basic proficiency in MATLAB and Simulink. Familiarity with linear algebra, system dynamics and elementary signal processing is also recommended to benefit fully from the course.
Before the course, participants are encouraged to complete the free online tutorials MATLAB Onramp and Simulink Onramp, provided by MathWorks.
Course Abstract
This advanced course introduces intelligent and adaptive methods for fault detection, isolation, identification and health monitoring in dynamic and autonomous systems. Starting from model-based residual generation and analytical redundancy, the course progresses through adaptive estimation, dynamic observers, filtering methods, neural and fuzzy models, and data-driven learning approaches. Particular attention is given to robustness against modelling uncertainty and disturbances, online adaptation, and integrating diagnostic information into supervisory and fault-tolerant control architectures. Practical engineering examples illustrate the methods and their relevance to autonomous systems, supported by MATLAB and Simulink implementations.
Learning Objectives and Expected Outcomes
The course aims to develop participants’ understanding of model-based, adaptive and learning-based methods for fault diagnosis, together with practical skills for their implementation and evaluation in MATLAB and Simulink. Particular attention is given to robustness, diagnostic performance, and the role of fault information in supervising autonomous systems.
Upon successful completion of the course, participants will be able to:
- Explain the principles of fault detection, isolation and identification, and distinguish their roles in condition monitoring and system supervision.
- Develop and evaluate residual generators based on analytical redundancy, dynamic observers and filtering methods.
- Apply recursive estimation methods to detect and assess parametric changes in dynamic systems.
- Use neural and fuzzy models for system modelling and fault diagnosis, recognising the importance of training and validation.
- Compare model-based and data-driven diagnostic methods in terms of sensitivity to faults, robustness to uncertainty and noise, and diagnostic performance.
- Implement and analyse selected diagnostic schemes in MATLAB and Simulink, and discuss how their results can support supervisory and fault-tolerant control.
Assessment and Self-Assessment
Participants will receive feedback during the practical examples and exercises. The course concludes with a final test consisting of 30 multiple-choice questions covering the main concepts and methods discussed in the lectures.
For self-assessment and preparation, participants can download a sample set of 10 multiple-choice questions (PDF) and a separate answer key (PDF).
Course Programme
Course duration: 8 lectures of approximately 2.5 hours each, for a total of 20 hours.
Week 1 — Model-Based, Robust and Adaptive Diagnosis
- Lecture 1: Introduction to Fault Diagnosis and Condition Monitoring
- Fault detection, isolation and identification
- Condition monitoring and the role of diagnosis in autonomous systems
- Fault models and engineering applications
- Introduction to the MATLAB and Simulink examples
- Lecture 2: Model-Based Fault Detection Methods I
- Analytical redundancy and residual generation
- Observer-based approaches
- Residual evaluation for fault detection and isolation (FDI)
- Practical examples in MATLAB
- Lecture 3: Model-Based Fault Detection Methods II
- Observer banks and unknown-input observers
- Parity-space methods
- Kalman filtering and diagnosis from noisy measurements
- MATLAB and Simulink exercises
- Lecture 4: Robust and Adaptive Fault Diagnosis
- Modelling uncertainty, disturbances and residual robustness
- Fault sensitivity and disturbance rejection
- Recursive least squares (RLS) with a forgetting factor for parametric fault diagnosis
- Online adaptation and MATLAB and Simulink implementation
Week 2 — Learning-Based and Integrated Diagnosis
- Lecture 5: Foundations of Data-Driven Fault Diagnosis
- Data-driven techniques for anomaly detection and fault diagnosis
- Signal processing and data preparation
- Training, validation and performance evaluation
- Practical session in MATLAB
- Lecture 6: Neural Models and Machine Learning for Fault Diagnosis
- Artificial neural networks and training methods
- Neural models for static and dynamic systems
- Support vector machines (SVMs) for fault diagnosis
- Practical examples and validation in MATLAB
- Lecture 7: Fuzzy Models and Advanced Data-Driven Diagnosis
- Fuzzy logic and fuzzy models for fault diagnosis
- Neural and fuzzy approaches to nonlinear system modelling
- Diagnostic applications and model validation
- Practical exercises using MATLAB and Simulink
- Lecture 8: Integrated Model-Based and Data-Driven Approaches
- Integration of learned models, dynamic observers and filters
- Comparative analysis and diagnostic performance evaluation
- Use of diagnostic information in supervisory and fault-tolerant control
- Practical workshop and simulations
Main Topics
- Fault detection, isolation and identification
- Fault modelling, analytical redundancy and residual generation
- Output observers, unknown-input observers and Kalman filtering
- Robustness to uncertainty, disturbances and measurement noise
- Recursive estimation and identification of parametric faults
- Neural and fuzzy models: training, validation and diagnostic applications
- Integration of model-based and data-driven methods
- Diagnostic performance evaluation and links to system supervision
- MATLAB and Simulink applications
Lecture Notes
- Part 1. Introduction to fault diagnosis, residual generation and evaluation, and robustness problems. Download Part 1 (PDF).
- Part 2. Recursive least squares with a forgetting factor for the detection of process and component faults. Download Part 2 (PDF).
- Part 3. Introduction to artificial neural networks, training methods and validation. Download Part 3 (PDF).
- Part 4. Neural models for dynamic systems and fault diagnosis. Download Part 4 (PDF).
- Part 5. Fuzzy logic and fuzzy models for fault diagnosis. Download Part 5 (PDF).
Application Examples in MATLAB and Simulink
Some examples were developed for earlier MATLAB and Simulink releases, as indicated below. Check the required release and toolboxes before running them; compatibility with your installed version may require adjustments.
- System Modelling and Fault Models. MATLAB script defining the process model; Simulink model for discussion; final Simulink model.
- Parametric Fault Generation in a Dynamic Process. Simulink model for data generation with parametric changes.
- Observer Bank Design for a Single-Input, Multiple-Output (SIMO) Model. Three dynamic observers for fault detection and isolation: MATLAB script for the process model and observer design; Simulink model with output sensor faults; Simulink model with input sensor faults; MATLAB script for residual evaluation.
- Recursive Estimation for Parametric Fault Diagnosis. RLS algorithm with a forgetting factor; MATLAB script for visualising the results; Simulink model for data generation with parametric changes; Simulink model for input and output fault generation.
- Recursive Estimation with a Forgetting Factor in Simulink. Simulink model using an RLS block (MATLAB R2016).
- Comparison of Recursive Estimation Methods for Parametric Fault Diagnosis. Simulink model comparing recursive estimates.
- Neural Network Training and Nonlinear Static Function Approximation. Examples using the legacy Neural Network Toolbox: data generation for neural network training; MATLAB script for training a multilayer perceptron (MLP) with backpropagation; Simulink validation model requiring the trained network; MATLAB script for evaluating network performance; Simulink validation example with the trained network included; Simulink example with a trained network and a step fault.
- Magnetic Levitation System Modelling and Fault Diagnosis. problem definition and supporting files (ZIP); worked solutions (ZIP).
References: Fundamental Monographs and Textbooks
- Isermann, R. (2011). Fault-Diagnosis Applications: Model-Based Condition Monitoring: Actuators, Drives, Machinery, Plants, Sensors, and Fault-tolerant Systems. Springer. ISBN: 978-3-642-12766-3. DOI: 10.1007/978-3-642-12767-0.
- Ding, S. X. (2008). Model-based Fault Diagnosis Techniques: Design Schemes, Algorithms, and Tools. 1st ed. Springer. ISBN: 978-3-540-76303-1. DOI: 10.1007/978-3-540-76304-8.
- Isermann, R. (2006). Fault-Diagnosis Systems: An Introduction from Fault Detection to Fault Tolerance. Springer. ISBN: 978-3-540-24112-6. DOI: 10.1007/3-540-30368-5.
- Blanke, M., Kinnaert, M., Lunze, J. and Staroswiecki, M. (2003). Diagnosis and Fault-Tolerant Control. 1st ed. Springer. ISBN: 3-540-01056-4. DOI: 10.1007/978-3-662-05344-7.
- Korbicz, J., Kowalczuk, Z., Kościelny, J. M. and Cholewa, W. (eds.) (2004). Fault Diagnosis: Models, Artificial Intelligence, Applications. Springer. ISBN: 978-3-540-40767-6. DOI: 10.1007/978-3-642-18615-8.
- Simani, S., Fantuzzi, C. and Patton, R. J. (2003). Model-based Fault Diagnosis in Dynamic Systems Using Identification Techniques. Springer. ISBN: 978-1-85233-685-1. DOI: 10.1007/978-1-4471-3829-7.
- Basseville, M. and Nikiforov, I. V. (1993). Detection of Abrupt Changes: Theory and Application. Prentice Hall. ISBN: 0-13-126780-9.
- Chen, J. and Patton, R. J. (1999). Robust Model-Based Fault Diagnosis for Dynamic Systems. Kluwer Academic Publishers. ISBN: 978-0-7923-8411-3. DOI: 10.1007/978-1-4615-5149-2.
- Chiang, L. H., Russell, E. L. and Braatz, R. D. (2001). Fault Detection and Diagnosis in Industrial Systems. Springer. ISBN: 978-1-85233-327-0. DOI: 10.1007/978-1-4471-0347-9.
- Gertler, J. (1998). Fault Detection and Diagnosis in Engineering Systems. Marcel Dekker. ISBN: 0-8247-9427-3.
- Hadjicostis, C. N. (2002). Coding Approaches to Fault Tolerance in Combinational and Dynamic Systems. Kluwer Academic Publishers. ISBN: 978-0-7923-7624-8. DOI: 10.1007/978-1-4615-0853-3.
- Liu, G. P. and Patton, R. J. (1998). Eigenstructure Assignment for Control System Design. Wiley. ISBN: 978-0-471-97549-6.
- Patton, R. J., Frank, P. M. and Clark, R. N. (eds.) (1989). Fault Diagnosis in Dynamic Systems: Theory and Applications. Prentice Hall. ISBN: 0-13-308263-6.
- Patton, R. J., Frank, P. M. and Clark, R. N. (eds.) (2000). Issues of Fault Diagnosis for Dynamic Systems. Springer. ISBN: 978-3-540-19968-7. DOI: 10.1007/978-1-4471-3644-6.
Further Reading on Data-Driven Methods
- Ding, S. X. (2014). Data-driven Design of Fault Diagnosis and Fault-tolerant Control Systems. Springer. ISBN: 978-1-4471-6409-8. DOI: 10.1007/978-1-4471-6410-4.
Additional Reading and Application Downloads
- Model-Based Fault Diagnosis for Industrial Processes. Extended report by Silvio Simani, October 2007. Download the report (PDF).
- Lecture Notes, Chapters 1 and 2. Extract from Silvio Simani’s extended report, October 2007. Download Chapters 1 and 2 (PDF).
- Model-based fault-detection and diagnosis — status and applications. Journal article by Rolf Isermann, 2005. Download the article (PDF).
- Output Observer Design for Fault Detection and Isolation. Example with measurement noise, developed for MATLAB 6.1. Download MATLAB and Simulink files (ZIP).
- Observer Bank Design for a SIMO Model. Three observers, with two MATLAB scripts and one Simulink model, developed for MATLAB 6.1. Download MATLAB and Simulink files (ZIP).
- Kalman Filter Design. State estimation for a model with noise, developed for MATLAB 6.1. Download MATLAB and Simulink files (ZIP).
- Kalman Filtering for Fault Diagnosis. Measurement noise, output sensor faults and statistical tests on residuals, developed for MATLAB 6.1. Download MATLAB and Simulink files (ZIP).
- Nonlinear Models and Neural Network Training. Download MATLAB and Simulink examples (ZIP).
- Integration of Neural and Fuzzy Models with Observers and Filters. Examples for fault detection and isolation. Download MATLAB and Simulink examples (ZIP).
Instructor Biography and Contact
Silvio Simani is a Professor of Automatic Control at the Department of Engineering, University of Ferrara, Italy. His research interests include fault diagnosis, fault-tolerant control, system modelling and identification, filtering, and neural and fuzzy methods for dynamic systems. He has authored journal articles, conference papers and books in automatic control and related engineering fields, and has extensive international teaching experience through doctoral courses, summer schools and specialised seminars.
Prof. Simani collaborates with academic institutions on international research and educational activities.
For questions or further information:
Email: silvio.simani@unife.it
Telephone / WhatsApp: +39 320 304 4076.