May 22, 2008

Postdoctoral Position in Statistics, Institut Francais du Petrole

Postdoctoral Position
Institut Français du Pétrole (IFP), Solaize, France

Subject : Response Surfaces under constraints

Because physical models are not predictive enough in motor calibration design, to represent engine exhausts, one has to resort to response surfaces techniques. Experimental data are fitted by means of non linear regression, locally when the torque and the engine speed are both fixed, or globally when everything vary.

To check the results, engineers rely on their expertise. For example, locally, CO2 emissions should decrease when the main injection timing (AVI) and the fuel injection pressure (PRAIL) increase. Globally, constraints coming from physical limitations and considerations become more and more stringent.

However, traditional regression techniques do not guarantee that all the constraints will be satisfied. In the case of CO2, one should observe monotony constraints, i.e. the response should increase or decrease monotonically when AVI and PRAIL vary. Similarly, maximum or minimum values are not guaranteed to be respected, since the calculated response can oscillate between the fitting points.

From the end of 90's, some new methods have been proposed to alleviate these limitations. A bibliographical review shows that many different techniques have been proposed, based on available regression techniques. Smoothing splines have been used, as well as generalized additive models, krigging or wavelets. Some of these developments are available through internet.

The goal of this post-doctoral study is to propose strategies to tackle these problems, to test and use the new techniques that have to be developed, and to apply them to the case of responses surfaces in the field of motor calibration. Coding will be done in MATLAB. Expected difficulties concern the high dimension of the regressors space, as well as the calculation time which has to stay manageable.

Education and qualification:

PhD in applied mathematics or in statistics, solid background in experimental design and response surfaces methodology. Practice of a programming language like MATLAB, C or FORTRAN.

Location: Institut Français du Pétrole (IFP, www.ifp.fr) at Solaize, near Lyon
Duration: 12 to 18 month
Salary: ~2862Euro /month (+ 474Euro /month for lodging expenses for foreigners coming from foreign universities only)

Application: Interested candidates should send a CV, a motivation letter, a publication list and referee letters to:

Dr. François WAHL, IFP
Francois.Wahlifp.fr

Please quote 10 Academic Resources Daily in your application to this opportunity!

If you want to receive academic resources in your e-mail on daily basis, please subscribe to 10resources-subscribe@yahoogroups.com.