A neural network approach for efficient calculation of the current correction value in femtoampere range for a new generation of ionizing radiation monitors at CERN

The European Organization for Nuclear Research (CERN) conducts experiments that involve colliding beams of particles either together or into stationary targets. During these interactions, stray radiation may be generated. The ionizing radiation detectors installed at several locations close to the beam lines and targets of these areas allow CERN radiation protection by precisely monitoring radiation levels. Radiation monitoring is one of the main responsibilities of the Radiation Protection Group and a crucial task to indirectly ensure safety at CERN and its surrounding environment. After 30 years of reliable service, the ARea CONtroller (ARCON) system has reached the end of its lifecycle. A new generation of radiation monitors called CROME (Cern RadiatiOn Monitoring Electronics) has been devel-oped at CERN. These monitors incorporate embedded processing capabilities in order to execute various algo-rithms, such as evaluation of the real electrical current generated by the radiation detectors when they are subject to ionizing fields. This paper presents a case study of a new method for offset correction of a femtoampere current. At this scale, the measured current is sensitive to surrounding environmental factors, such as temper-ature, vibration. and the permittivity of the air. To guarantee the high precision of calculation and real-time operation, and to overcome the limitations of the field-programmable gate array (FPGA) platform used, a novel method utilizing a neural network approach is proposed. The results obtained with a new model are very satisfactory in terms of both accuracy of prediction and reduced computational complexity. This may encourage further usage of neural networks in safety-critical systems