Advanced cryogenic process control and monitoring for the SPIRAL2 superconducting LINAC

TL;DR

This paper discusses advanced cryogenic process control and monitoring techniques for the SPIRAL2 superconducting LINAC, combining physical models and machine learning to optimize cryogenic stability under high heat loads.

Contribution

It introduces a novel observer for dynamic heat load estimation based on thermodynamic models and machine learning, enhancing cryogenic control.

Findings

Maintains cryogenic conditions under high heat loads

Demonstrates effectiveness of model-based control in operational settings

Develops a virtual heat load observer using combined approaches

Abstract

SPIRAL2 is a superconducting accelerator for protons, deuterons and heavy ions delivering a maximum beam power of 200 kW at 40 MeV (for deuteron beams). 26 superconducting quarter wave cavities are operated at 4.4 K, plunged in a liquid helium bath with a drastic phase separator pressure control. Previous years have seen the development of advanced process control for cryogenics allowing to cope with high heat load dynamics thanks to an automatic heat dissipation compensation and a model based control. The latter is based on models, using the Simcryogenics library, optimized and linearised in the Programmable Logic Controllers. The SPIRAL2 operation has demonstrated that such control allows to keep the specified conditions for RF and beam operation even at levels of heat load dissipation approaching the physical limits of the cryogenic system. These developments allowed to synthesise a virtual observer of the dynamic heat load dissipated by the cavities. The present paper summarises the development of such observer based on the physical thermodynamic model and on machine learning techniques.