ERNSI Workshop Cambridge,UK

September 16-19, 2001



 



The meeting will be held at  Pembroke College, Cambridge, September 16-19, 2001.

The University of Cambridge is rich in history - its famous Colleges and University buildings attract visitors from all over the world. But the University's museums and collections also hold many treasures which give an exciting insight into some of the scholarly activities, both past and present, of the University's academics and students.  The University of Cambridge is one of the oldest universities in the world and one of the largest in the United Kingdom. Its reputation for outstanding academic achievement is known world-wide and reflects the intellectual achievement of its students, as well as the world-class original research carried out by the staff of the University and the Colleges.
 


Travel Information

Train Timetables:
Cambridge to London  and  London to Cambridge.

Airports (Stansted, Luton, Heathrow and Gatwick).

Various links to low price airlines:
 GoEasy JetBuzzRyanair .

There are buses departing Stansted, Luton, Heathrow and Gatwick Airports just about every hour.  For exact timetables you can look up your journey on the  National Express website.  There is also a  train  from Stansted to Cambridge which is quite inexpensive (6 pounds one way).

Information on Taxis in Cambridge.

Map of  Cambridge.

Cambridge weather.
 

Arrival

When you arrive, please report to the Pembroke College Porters' Lodge, which is located at the intersection of Trumpington Street and Pembroke Street. They will give you a key to your room.

Please note that the first meal arranged for the Workshop is Sunday dinner in the evening. That means that those arriving on Saturday will have to make their own meal arrangements. This should not be a problem, as there are several hundred restaurants, pubs, etc to choose from in Cambridge.

You will be asked to vacate your room on the morning of your day of departure, but facilities for storing luggage will be available.

Please note:  You will need to vacate your rooms by 9.30am on day of departure.  Luggage can be left at the Porters Lodge if necessary.
 


Registration

Team leaders have been sent a request for details of those attending, to be returned by Friday, August 31st if possible and no later than Friday, September 7th.

Payment for workshop participants will be collected from Team Leaders, not from individual participants.

NOTE! The ERNSI-meeting is open only to participating groups in the EU TMR Project System Identification.


Programme

All activities in the Old Library, unless otherwise stated.
 
Sunday 16 September Monday 17 September Tuesday 18 September Wednesday 19 September
9.00 Jan Maciejowski, Welcome and opening remarks Regular talks session : 
Talk 4 Scherrer 
Talk 5 Markovsky and De Moor 
Talk 6 D Trummer and A Al-Falou
Tutorial Lecture:
A new look at the state representation problem 
- Jan C. Willems
9.15 Prof. David Sandoz (Control Technology Centre Ltd  and Manchester Univ) - An Initial Appreciation of the suitability of Multivariate Statistics to enhance capability for Condition Monitoring of Industrial Processes
9.30
10.00
10.30 Tea/Coffee Break Tea/Coffee Break Tea/Coffee Break
11.00 Regular talks session: 
Talk 1 T Ribarits
Talk 2 L Johnston and M Gevers
Talk 3 E. Fabre, A Benveniste, S. Haar

11.00-12.30 Task Group 3&4
Regular talks session:
Talk 10 J Schoukens, R Pintelon, T Dobrowiecki
Talk 11 X Bombois, M Gevers, G Scorletti
Talk 12 Q Zhang
11.30
12.00
12.30 Lunch (in Hall
Separate lunch for team leaders (Christopher Smart Room, E5)
(Lunch table for Ph.D. students and for post-docs)
Lunch in Hall
Lunch in Hall
1.00
1.30
2.00 2.00-2.30 Oral Presentations (2 mins per poster)
 

2.30-4.00 Task Group 1&2

Regular talks session: SI and signal processing
Talk 7 Jan van Schuppen
Talk 8 Byrnes, Enqvist, Lindquist
Talk 9 I Ioslovich, I Seginer, PO Gutman
2.30 Collaboration planning (+ second sessions of Task Groups?)
3.30
3.45 Guided Tour of  Fitzwilliam Museum
4.00 Tea/Coffee Break Tea/Coffee Break. CLOSE
4.30 Software demonstrations:
1) Kerrigan, Chen, Maciejowski
2) Ljung and Zhang
5.00 To be decided (Social)
5.30 Posters, individual software demos
6.00
6.30 Free discussion, small groups etc.
7.00
7.30 Dinner in Hall Dinner in Hall Special Dinner in Old Library
8.00
8.30 Posters mounted on display boards
9.00 Introductions
9.30

Task Groups (Monday 2.00-4.00 and Tuesday 11.00-12.30)
Mon 2.00-4.00 Task Group SI for nonlinear systems - Jan van Schuppen -Chair (Room N7)
                      Task Group SI for control - Michel Gevers - Chair (Old Library)

Tues 11.00-12.30 Task Group SI for linear systems - Lennart Ljung - Chair (Old Library)
                          Task Group SI for signal processing - Jean-Jacques Fuchs (Room N7)

Regular Talks Sessions (Each Talk: max. 20 minutes + 10 minutes discussion)
Monday 11.00-12.30
Talk 1 - Parametrization of Linear Systems by Data Driven Local Coordinates - Thomas Ribarits (TUW)
Talk 2 - Model reduction versus direct low order identification: a variance analysis. - Leigh Johnston and Michel Gevers (CESAME - UCL)
Talk 3 - The state estimation problem for DEDS -  E. Fabre, A Benveniste, S. Haar (IRISA)

Tuesday 9.00-10.30
Talk 4 - The local optimality of stochastically balanced truncation - Wolfgang Scherrer (TUW)
Talk 5 - Multi-Model System Parameter Estimation -  Ivan Markovsky, Bart De Moor (ESAT-SISTA, K.U.Leuven)
Talk 6 - Identification of recurrent neural nets - Dietmar Trummer (TUW)

Tuesday 2.00-3.30
Talk 7 -  Problems of realization and control of cell reaction networks - Jan van Schuppen (CWI)
Talk 8 - Identifiability and well-posedness of linear models in terms of cepstral coefficients, covariance lags and Markov parameters -Byrnes, Enqvist and Lindquist (KTH)
Talk 9 - Dominant parameter selection in the marginally identifiable case - Ioslovich, Seginer and Gutman (KTH)

Wednesday 11.00-12.30
Talk 10 - Some thoughts on the stability of closed loop systems in the presence of nonlinear distortions - Johan Schoukens, Rik Pintelon, and T. Dobrowiecki (Vrije Universiteit Brussels)
Talk 11 - Controller validation for stability and performance based on a frequency domain uncertainty region obtained by stochastic embedding - X. Bombois (TU Delft),
              M. Gevers (UCL, Louvain-la-Neuve), G. Scorletti (LAP ISMRA, Caen)
Talk 12- Adaptive Observer for MIMO Linear Time Varying Systems - Qinghua Zhang (IRISA)
 

POSTERS (Space for each poster: 1.2m high, 1m wide)

1 - Identification for control: optimal input design with respect to a worst-case $\nu$-gap cost function - Roland Hildebrand (UCL)
2 - On Iterative Feedback Tuning for non-minimum phase plants - A.Lecchini and M.Gevers (UCL)
3 - Regressor selection with the Analysis of Variance Method - Ingela Lind (ULIN)
4 - Use of multiple models in predictive control - Thomas Schon (ULIN), Jan Maciejowski (UCAM) and Lennart Ljung (ULIN)
5 - A statistical approach to system identification and fault detection in a static system - Anna Hagenblad (ULIN)
6 - Nonlinear identification and control of hydraulic actuators in variable cam timing engines - Umut Genc (UCAM)
7 - Global Identifiability Analysis using Algorithms for detecting Connected Semi-Algebraic Components - Dorina Jibetean (CWI)
8 - Variance properties of a two-step ARX estimation procedure - F.Tjärnström and Ljung (ULIN)
9 - Some benefits of aliasing in time series analysis -S. de Waele and P. Broersen (CWI)
10 - First estimates of Wiener-Hammerstein systems using a random multisine excitation - P. Crama (VUB)
11- Fuzzy Model Validation Using the Statistical Local Approach - G. Rigatos and Q. Zhang (IRISA)
12 - Variance-bias tradeoff in Finite Impulse Response estimates obtained by correlatoin analysis - Martin Enqvist
*13 - Using MILP/MIQP for identification of piecewise affine systems - Jacob Roll (ULIN)

NOTE : * Subject to available space

SOFTWARE DEMOS

Linear and Bilinear Identification Toolbox - Eric Kerrigan, Huxin Chen and Jan Maciejowski (UCAM)

Nonlinear Toolbox - Lennart Ljung (ULIN) and Qinghua Zhang (IRISA)
 

ABSTRACT - GUEST SPEAKER

An Initial Appreciation of the suitability of Multivariate Statistics to enhance capability for the Condition Monitoring of Industrial Processes.
- D.J. Sandoz, Control Technology Centre Ltd.

In 1997, Invensys acquired Predictive Control Ltd and the associated software “Connoisseur”. Predictive Control Ltd was founded in 1988 as a spin off company from the University of Manchester and Connoisseur is a product for the design and application of Industrial Model Predictive Controllers.

Predictive Control Ltd. had established a programme to undertake the development of a new software product to provide a capability for process condition monitoring to be exploited in its own right or as a companion with Connoisseur. This product (tentatively known as Perceptor) has continued to be developed at Manchester University within Control Technology Centre Ltd, a company wholly owned by the University. Trials are now underway with a number of Industrial Companies to evaluate capability. These trials are a penultimate stage in a progression to introduce Perceptor to the market place.

Multivariate Statistics (for example Principal Component Analysis and associated Probability Density Function statistics) has been and remains a major topic for academic research. It is argued that it provides the foundation for pragmatic multivariate approaches for the condition monitoring of industrial processes. However, the reality is that there has been minimal real-time exploitation with industrial plant, to the extent that it is questionable that there is even a single instance of successful exploitation that is in continuous operation anywhere in the world. This is in clear contrast to the success of the multivariate control engineering methods, which have now become a commodity in industry.

This presentation is concerned with the capability of Perceptor, including methods to address issues of condition monitoring and classification in both Continuous and Batch processes. The results of early exposures to industrial data will be reviewed. The emphasis in the presentation is to provide an initial feel for the potential of Perceptor to contribute beneficially in real-time to the operation of industrial processes.
 
 

HALL
 


 
 
 

Entrance to Old Library