Deep learning for 2D Fluorescence. Data processing optimization on black plastics for classification/quantification dedicated to recycling applications
Direction Expérimentation Procédés

IFP Energies nouvelles (IFPEN) est un acteur majeur de la recherche et de la formation dans les domaines de l’énergie, du transport et de l’environnement. De la recherche à l’industrie, l’innovation technologique est au cœur de son action, articulée autour de quatre priorités stratégiques : Mobilité Durable, Energies Nouvelles, Climat / Environnement / Economie circulaire et Hydrocarbures Responsables.

Dans le cadre de la mission d’intérêt général confiée par les pouvoirs publics, IFPEN concentre ses efforts sur :

• l’apport de solutions aux défis sociétaux de l’énergie et du climat, en favorisant la transition vers une mobilité durable et l’émergence d’un mix énergétique plus diversifié ;
• la création de richesse et d’emplois, en soutenant l’activité économique française et européenne et la compétitivité des filières industrielles associées.

Partie intégrante d’IFPEN, l’école d’ingénieurs IFP School prépare les générations futures à relever ces défis.

Deep learning for 2D Fluorescence. Data processing optimization on black plastics for classification/quantification dedicated to recycling applications

IFPEN works on developing recycling processes especially for plastics. Solutions for characterizing the incoming flux of solid materials rely a lot on hyperspectral imaging (usually based on VIS-NIR reflected light). However, from a black object all the light is adsorbed and not much is reflected. Fluorescence spectroscopy on the other hand measures the emitted light of a sample after excitation at a specific wavelength. 2D fluorescence scans through different excitation wavelengths and different emission wavelengths.

Its signals are very rich nonlinear spectroscopic data able to describe sensitively samples.

A fluorophoressing molecule’s spectrum will be strongly influenced by its close environment (pH, viscosity, Temperature, …).

Classical chemometric tools available for exploiting this rich data are not satisfactory (Parafac and nPLS). They are best for linear signals.

Manual Variable selection based on data observations proved quite effective in exploiting this type of data, but more systematic tools would be appreciated.

Deep learning strategies can be investigated to help:

• Explore the full potential on these extensive 2D signals;
• Extract the most descriptive and robust variables for characterizing the dataset.

The applications targets would be:

• to classify objects by making the most of the signal;
• to predict quantitatively contaminations levels in a sample.

Description

The job would consist in optimizing the global acquisition/data processing protocol.

List of tasks:

• Task 1: acquire signals for the database (with the help of qualified laboratory personnel):Statics 2D signals and optimized Dynamic signals that mimic online measurements.
• Task 2: processing optimization of signal and from the database calibrate and validate models using Deep Learning and Machine Learning for exploratory and quantitative work.
• Task 3: report and oral presentations

The work will also include a literature review of fluorescence data processing, with a focus on recent deep learning approaches.

Required profile

Engineer or university Master level M2

• English or French, curiosity, enthusiasm, autonomy. 
• Strong background in data processing.
• Knowledge in chemical analysis, engineering and polymers would be a plus.