Time-Delayed Koopman Network-Based Model Predictive Control for the FRIB RFQ

TL;DR

This paper introduces a deep learning-based predictive control method using a Koopman network with LSTM to effectively manage RFQ resonance frequency detuning at FRIB, addressing delays and nonlinearities in the system.

Contribution

It develops a novel LSTM-based Koopman network model for predicting RFQ frequency behavior and integrates it into a model predictive control framework for improved frequency regulation.

Findings

The model accurately predicts RFQ resonance frequency using historical data.

The proposed MPC reduces control time by 50% compared to PID control.

Deep learning-based MPC achieves precise and rapid frequency control.

Abstract

The radio-frequency quadrupole (RFQ) at the Facility for Rare Isotope Beams (FRIB) is a critical device to accelerate heavy ion beams from 12 keV/u to 0.5 MeV/u for state-of-the-art nuclear physics experiments. Efficient control of the RFQ resonance frequency detuning still remains a challenge because the temperature-sensitive frequency is solely control by a cooling water system, exhibiting complicated transport delay and nonlinearity in the heat transfer processes. In this work, we propose a long-short term memory (LSTM)-based Koopman network model that can simultaneously learn the time-delayed and non-delayed correlations hidden in the historical operating data. It is proven that the model can effectively predict the behavior of the RFQ resonance frequency using historical data as inputs. With this model, a model predictive control (MPC) framework based on the Newton-Raphson method is proposed and tested. We demonstrate that the MPC framework utilizing deep learning model is able to provide precise and rapid control for the RFQ frequency detuning, reducing the control time by half compared to the proportional-integral-derivative (PID) controller.