Speaker
Description
The Multi-Objective Genetic Algorithm (MOGA) is a powerful and increasingly adopted method for optimizing both linear and nonlinear beam dynamics in accelerator lattices, particularly for ultralow-emittance storage rings. Key objectives in this optimization include minimizing beam emittance for high brightness, maximizing dynamic aperture to ensure efficient particle injection, and expanding momentum aperture to enhance beam lifetime. However, these objectives are constrained by various strict lattice parameters, and magnet strengths which must be satisfied for a physically viable machine design. Integrating Machine Learning (ML) techniques with MOGA offers a promising solution to accelerate the optimization. By training ML models on the objective values that meet predefined constraints throughout MOGA’s evolution, we can predict optimal variable sets more efficiently. This ML-assisted MOGA approach not only speeds up the optimization process but also improves convergence towards feasible solutions. Our current work applies this combined ML and MOGA optimization approach to MAX 4U, an upgrade of the MAX IV 3 GeV ring, a state-of-the-art ultralow-emittance storage ring, with the goal of achieving sub-100 pm·rad emittance.
