Submission #238

Submission information
Submitted by Anonyme (not verified)
Tue, 07/18/2017 - 08:10
PhD on mathematical methods for the systematic identification of chemical compounds from complex spectrometric data
PhD Thesis
Ecole Centrale de Nantes / Total
Nantes / Lyon
Lubricants are used in many applications (car engines, gear boxes, etc.) to reduce friction between surfaces. Total is investing in advanced spectrometric equipment to obtain very detailed characterizations of the products. The techniques currently applied provide measurements in the form of high volume mono- or bidimensional spectra. The use of automated treatment methods is necessary to extract relevant information hidden in these complex data and to quickly identify the components present in a product based on their spectral fingerprint.

The analysis of multi-compound chemical substances based on spectrometric measurements can be formulated as a signal processing problem known as source separation or sparse representation. The feasibility of the application of these methods to lubricant data has been demonstrated through preliminary studies in the case of monodimensional spectra. However many aspects need to be investigated in ordre to build a tool able to identified components in a reliable manner. In particular, the bidimensional aspect of the data will need to be exploited to better discriminate chemincally similar compounds. The algorithms developed during the Ph.D. will be implemented in a software tool designed for laboratory technicians.
Tue, 08/15/2017