# Automatic tuning of Free Electron Lasers

## TL;DR

This paper presents an automatic tuning system for Free Electron Lasers that combines empirical and model-based methods, improving stability and reducing manual effort at facilities like FLASH.

## Contribution

It introduces a hybrid automatic tuning approach for FELs, integrating empirical control with potential machine learning enhancements, demonstrated at FLASH.

## Key findings

- Achieved automatic SASE tuning at FLASH
- Reduced manual tuning time and improved stability
- Discussed future integration of machine learning techniques

## Abstract

Existing FEL facilities often suffer from stability issues: so electron orbit, transverse electron optics, electron bunch compression and other parameters have to be readjusted often to account for drifts in performance of various components. The tuning procedures typically employed in operation are often manual and lengthy. We have been developing a combination of model-free and model-based automatic tuning methods to meet the needs of present and upcoming XFEL facilities. Our approach has been implemented at FLASH \cite{flash} to achieve automatic SASE tuning using empirical control of orbit, electron optics and bunch compression. In this paper we describe our approach to empirical tuning, the software which implements it, and the results of using it at FLASH. We also discuss the potential of using machine learning and model-based techniques in tuning methods.