New job offer job_b0e0dc497ea7bbc133f75af204e18630
Title: CNN neural network – managing data uncertainty in the learning database.
Date: 2021-09-24 07:50
Job_Duration: 12 mois
The postdoctoral fellow will join an internal project shared between different CEA entities. The aim is to develop algorithms able to take into account the uncertainty in the learning database of neural networks. Indeed, the database can contain simulated data as well as real data. By definition, the simulated data is impacted by the uncertainty of the model used for the calculation and the real data is impacted by the uncertainty of the measuring instrument. In this postdoctoral research, we focus only on the uncertainty of real data and we suppose that the simulated data is exact. The aim is to take into account such information in the building process of the neural network.
The project fits into the context of the dynamic state estimation of liquid-liquid extraction and benefits of its knowledge-based simulator as well as industrial data. Indeed, the status of an industrial chemical process is accessible through operating parameters and available monitoring measures. However, the measures being inherently associated with uncertainty, it is necessary to make the data consistent with process knowledge. Therefore, the goal is to find the best data set of operational parameters (input of the knowledge-based simulator) to provide the model to estimate the real process state known through monitoring measures (output of the knowledge-based simulator). A convolutional neural network (CNN) is being developed in another postdoctoral project to solve the inverse problem to find the best input thanks to the measured output. A consistent set of operating parameters is going to be obtained and state of the process is going to be known during the dynamic regime of the liquid-liquid extraction process.
To build the CNN, data will be obtained from the knowledge-based model and from measuring instruments. As said above, the simulated data will be considered exact. The goal of this postdoctoral project is to take into account the uncertainty of the measured data into account during the learning step of the CNN.
This first step is to evaluate the impact of the uncertainty of operational parameters on the outputs of the knowledge-based model. This step will need to connect the knowledge-based model to URANIE, internal platform developed by CEA ISAS. Uranie is a plateform based on the ROOT data analysis software, developed by CERN. URANIE notably displays algorithms allowing the propagation of uncertainties and sensitivity analyzes. The results obtained with propagation of uncertainties technics will be compared with the measures of the output data and their own uncertainties. .
This knowledge must be taken into account in the second part of the project. The uncertainty observed on the outputs should be taken into account in the learning loop to improve the estimation of the operational parameters by the CNN. The impact of these uncertainties on the CNN computed results must be assesed in order to trust the ability of the CNN to estimate the state of the process.
Through this project, we are at the heart of the thematic of digital simulation for the best control of complex systems. This post-doc is an opportunity to build a tool to help the operation of a complex process by using the expertise of the process and modern techniques of artificial intelligence, which will be able to take into account the volume and “real time” requirements.
 B. Dinh, M. Montuir et P. Baron, «PAREX, a numeric code for process design and integration.,» chez Global 2013, Salt-Lake City, 2013
 A. Duterme, M. Montuir, B. Dinh, J. Bisson, N. Vigier, P. Floquet et X. Joulia, «Process state estimation based on data reconciliation coupled with simulator » chez SFGP2019, Nantes, France, 2019.
 B. Dinh, V. Vanel, C. Sorel, A. Duterme, M. Montuir, G. Ferlay, « Process Simulation Tools for Process Control and Safeguards Purpose », International Nuclear Fuel Cycle conference, Global 2019, Seattle, Washington, USA, proceedings vol.1, pp 313-317
 F. GAUDIER, «URANIE: The CEA/DEN uncertayinty and sensitivity platform.» Procedia - Social and Behavioral Sciences, Volume 2, n°1, Issue 6, pp. 7660-7661, 2010.