| Home |
| General Info |
| Overview |
| People |
| Visitor Info |
| Links |
| Impressum |
| Research |
| Topics |
| Publications |
| Awards |
| Education |
| Courses |
| Thesis Projects |
| eLearning |
| Bulletin Board |
| Student Exchange |
| More... |
| News |
| Seminars |
| Events |
| In the Press |
| Jobs |
Lecturer/Time/Place/Prerequisites/Synopsis/Program/Contact/Sign Up |
Techniques in Process Monitoring and Diagnosis
- A Short Course -
November 28 - December 4, 2002 (12 lectures + exercises)
Sign Up (desired but noncommittal)Data files for exercises
Professor Ahmet Palazoglu
Department of Chemical Engineering and Materials Science
University of California
Davis, CA 95616 USA
anpalazoglu@ucdavis.edu
http://www.chms.ucdavis.edu/research/web/pse/ahmet/
On sabbatical leave at
Institute for Systems Theory in Engineering
University of Stuttgart
Pfaffenwaldring 70550 Stuttgart, Germany
palazoglu@ist.uni-stuttgart.de
Department of Chemical Engineering and Materials Science
University of California
Davis, CA 95616 USA
anpalazoglu@ucdavis.edu
http://www.chms.ucdavis.edu/research/web/pse/ahmet/
On sabbatical leave at
Institute for Systems Theory in Engineering
University of Stuttgart
Pfaffenwaldring 70550 Stuttgart, Germany
palazoglu@ist.uni-stuttgart.de
Time and Place
November 28 - December 4, 2002Daily lectures: 9:45-11:15 and 11:30- 13:00, additional exercises on two afternoons (probably on the 2nd and 4th day)
First meeting: November 28, 9:30, IST, Pfaffenwaldring 9, 2nd floor
Prerequisites
Basic StatisticsThe lecture will be given in English.
Synopsis
Following a brief review of basic statistical concepts, classical statistical process control tools such as Shewhart and EWMA charts are reviewed for comparison with advanced methods that are more appropriate for multivariable processes. Several techniques such as principal components analysis (PCA), partial least squares (PLS), Wavelet analysis, and Hidden Markov Models (HMM) are introduced to provide mathematical background for monitoring tasks. PCA and PLS based methods are presented for monitoring multivariable continuous processes, while contribution plots are used for assisting fault diagnosis. Trend detection using qualitative signal representation and HMM classification, wavelet-based hidden Markov trees for state estimation and trend classification, and multidimensional classification problems are also included.Techniques will be illustrated by hands-on exercises using MATLAB.
Program
November 28 – December 4, 2002Day 1: Introduction to Statistics, univariate charts (Shewhart, CUSUM, etc.), autocorrelation (Room 9.31)
Day 2: Multivariate statistics, Principal Component Analysis (PCA), fault diagnosis techniques
Day 3: Regression methods, Partial Least Squares (PLS), sensors
Day 4: Wavelets and signal filtering, Markov processes
Day 5: Fault diagnosis and classification using wavelets and hidden Markov models (HMMs)
For additional information and/or a desired but noncommittal sign up please contact: Thomas Eißing
Universität Stuttgart
|