Nonlinear optimization for a low-emittance storage ring.

Area of Science:

• Computational Science
• Optimization Algorithms
• Particle Accelerator Physics

Background:

• Multi-objective genetic algorithms (MOGA) are effective global optimization tools.
• MOGA performance is sensitive to the crossover parameter (ηc), requiring extensive computation for tuning.
• Existing multi-generation Gaussian process optimization (MGGPO) methods also benefit from parameter optimization.

Purpose of the Study:

• To introduce a self-adaptive crossover parameter strategy for MOGA.
• To integrate this adaptive strategy into MGGPO for improved trial solution generation.
• To evaluate the enhanced MOGA and MGGPO performance in nonlinear optimization tasks.

Main Methods:

• Implemented a self-adaptive crossover parameter that adjusts ηc each generation within MOGA.
• Applied the same adaptive strategy to MGGPO for parameter selection during trial solution production.
• Compared the performance of the new adaptive MOGA and MGGPO against their conventional counterparts.

Main Results:

• The self-adaptive strategy significantly improved the performance of both MOGA and MGGPO.
• Enhanced MOGA and MGGPO demonstrated superior results in nonlinear optimization.
• Specifically, improved performance was observed in the design of low-emittance storage rings.

Conclusions:

• Self-adaptive crossover parameters offer a computationally efficient and effective enhancement for MOGA and MGGPO.
• This adaptive approach leads to better optimization outcomes, particularly for complex engineering designs like storage rings.
• The proposed method reduces the burden of manual parameter tuning, saving significant computing time.