iten

Singularity

The aim of the “Singularity” project is to investigate the feasibility of a complete automation of an accelerator facility through the combined development of artificial intelligence (AI) software applications and a safety hardware able to control and monitor accelerator’s devices.

The proposed project aims to create hardware and develop software that allows the integration of artificial intelligence (AI) for risk management and the automated control of entire particle accelerator infrastructures.

In this context, the software development will focus on a dedicated AI-based Middle Layer for the accelerator control system in order to drive user requests through the control framework backend, operate the equipment and compute data analysis.

The main purpose of this layer is to enable advanced computer-literate end-users make basic modifications to the system without facing problems caused by the interaction with hardware front-end of the control system.

Such layer will host different classes of automated high-level applications (Operation, Conditioning, Beam Diagnostics and Fault Detection) through different classes of Artificial Intelligence strategies as Deep Learning, Clustering and Reinforcement Learning.

3-tier generic control system architecture with AI-based Middle Layer

About hardware development, an FPGA-based device will be developed by integrating, beyond the state-of-the-art, the main safety standards used in industrial and nuclear environments in both hardware and software designs.

This versatile device can be scaled to meet accelerator demands and, thanks to the high level of reliability, could be the first one able to operate as Machine Protection System (MPS) and as Personnel Safety System (PSS), to protect accelerator equipment and workers from prompt ionizing radiation exposure risks.

 

Risk reduction scheme through integration of safety standards from personnel and machine safety up to normal operation.

All the development, both hardware and software, will be modular and machine-independent so that it can be adapted in the most versatile way to fit any facility demand in terms of performance, architecture and scope of work (e.g. linac, storage-rings, user-facility, medical and industrial facility, etc…) in order to be scaled and composed as needed. INFN will benefit from a better management of the risks and costs of the facility, optimizing the accelerator up-time for operations and users experience.

Latest modified: 26 October 2020