Surrogate Modeling of the CLIC Final-Focus System using Artificial Neural Networks

TL;DR

This paper develops an artificial neural network-based surrogate model for the CLIC final-focus system, enabling faster optimization of beam parameters by replacing costly simulations with a trained predictive model.

Contribution

The paper introduces a neural network surrogate model for the CLIC final-focus system, facilitating efficient optimization of sextupole alignment parameters.

Findings

The surrogate model accurately predicts luminosity and beam sizes from sextupole offsets.

The model accelerates the optimization process compared to traditional simulation methods.

Effective in guiding parameter adjustments for improved collider performance.

Abstract

Artificial neural networks can be used for creating surrogate models that can replace computationally expensive simulations. In this paper, a surrogate model was created for a subset of the Compact Linear Collider (CLIC) final-focus system. By training on simulation data, we created a model that maps sextupole offsets to luminosity and beam sizes, thus replacing computationally intensive tracking and beam-beam simulations. This model was then used for optimizing the parameters of a random walk procedure for sextupole alignment.