Bochum 2015 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

P: Fachverband Plasmaphysik

P 6: Poster Session - Diagnostics

P 6.18: Poster

Monday, March 2, 2015, 16:30–18:30, Foyer Audimax

Acceleration of Bayesian Model Based Data Analysis through Software/Hardware — •Humberto Trimiño Mora — Max Plank Institut für Plasmaphysik, Greifswald, Germany

To control next generation of leading fusion experiments an improvement in machine control and plasma stability has to be reached in order to elevate fusion to a stage of effective operational energy source. This machine control improvement can be achieved by a betterment of signal processing from diverse diagnostics. Current trends in data analysis commonly focus on doing direct signal processing and analysis of a measured voltage or current to obtain the parameter of interest. But the possibilities of improving our estimations increase when we can use a technique that eases the process of incorporating our knowledge, or lack of it, into our way of analyzing the data. Model based data analysis can allow us to make a better estimation of the values of interest by considering what we know of the data and effectively introducing it into the estimation process through forward modeling and Bayesian inference but usually taking long processing times which make it not yet useful for real time processing. This project aims to use Bayesian probability theory and forward modeling to achieve a purely mathematical model through software/hardware, thus having a more informed estimation of a value and a rigorous determination of its uncertainty on a real-time frame. This projects reach includes not only application in the discussed area of interest, but also improvement of data integration and signal processing on smart systems or other platforms today that have incoming data from several peripherals.