Mainz 2014 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 94: DAQ, Trigger, Elektronik 4

T 94.7: Vortrag

Mittwoch, 26. März 2014, 18:20–18:35, GFH 01-731

The Neural Network Z Vertex Trigger for the Belle II Detector — Fernando Abudinen and •Sebastian Skambraks for the Belle II collaboration — Max-Planck-Institut für Physik (Werner-Heisenberg-Institut), Föhringer Ring 6, 80805 München

Triggering is crucial for the proper operation of particle physics experiments since the data rate generated by all the detector components is too large to be recorded for a later offline analysis. Studies for the Belle II detector showed the significant improvements on the z-vertex coordinate that can be achieved using Multi Layer Perceptrons (MLP), a sophisticated neural network approach. The z-coordinate of the vertex position would be a perfect criterion to discriminate signal and background. However, the experience from the Belle detector indicates that only a very low accuracy (ca. 10 cm) can be achieved with the classical trigger methods. By using the hit states and drift times (distance measures of charged tracks to the wires) from the Central Drift Chamber (CDC) the MLP is capable to provide an accuracy of ca. 1 cm for geometrically restricted areas of the detector. Therefore, a set of MLPs (about 2 millions) is needed. Since each MLP is specialized to a small phase space element, a preprocessing step is required to select the specialized MLPs. The Application of a preprocessing step together with the MLP ensemble could then cover the full CDC acceptance region of the Belle II detector.