MRADSIMIDE
Improving data analysis and reconstruction capabilities
Background
BEAMIDE is an innovative spin-off of INFN (Italian National Institute for Nuclear Physics) with core business of radiation hardness assurance services tests and software & innovative radiation detectors. The MRADSIMIDE pilot is about the development of a software set to simulate the radiation effects on electronic and electromechanical systems/subsystems (DUT). MRADSIM is a cross-platform system using OpenCascade, Qt5, CMake etc. and it provides a user-friendly interface to the users which are not experts in computing to do their simulations in autonomy.
Challenge
MRADSIM needs CPU, GPU, and disk space, but most of all training on how to implement algorithms for combined CPU and GPU usage and optimisation, to be able to simulate very large number of radiation events (e.g. 10^16 particles through a big GEO satellite over 20 years of activity) and
to use AI and ML techniques to run lesser number of events while approaching to the same result accuracy as in full run case.
EOSC services used
The Pilot used EGI Cloud Compute services through EGI-ACE, ROHub and API from RELIANCE and the expertise on AI and ML from AI4EOSC.
“
Results
The programmes were uploaded to test the first approaches focused to implement the superresolution issue, namely, to improve the 3D dose distribution obtained with a limited number of simulated events. The algorithms have been tested on EGI infrastructures, allowing to select the correct algorithm among many available. Multithread, multi-CPU computing in MRADSIM and in radiation detector’s data
acquisition, reconstruction and display codes have been implemented. Algorithms with AI/ML resulted in better images in MRADSIM 3D reconstructions for Dose-Depth Studies, and in GamCam (a crystal scintillator coupled to a Hamamatsu CHALLENGE SIOPM array).
Impact
TRL for the space module improved from TRL6 to TRL7.
Supported by
