ANL, Argonne, USA
Optimization through simulation is a time-consuming task in accelerator design, especially for high dimensional problems. We explored several parallel optimization techniques, including Parallel Genetic Algorithm (PGA), Hybrid Parallel Simplex (HPS), and Parallel Particle Swarm Optimization (PPSO), to solve some real world problems. The serial simplex method in elegant was used as a benchmark for newly-developed parallel optimization algorithms in Pelegant. PGA and HPS are not faster than the serial simplex method, but they more reliably find the global optimum. PPSO is well suited for parallel computing, allowing significantly faster turn-around given sufficient computing resources. Parallel optimization implementations in Pelegant thus promise to not only make optimization results more reliable, but also open the possibility of fast, "real time" optimization of complex problems for accelerator operation.
ANL, Argonne, USA
A method that tracks particles directly through a general magnetic field described in a 3D field table was added to the code elegant recently. It was realized by converting an arbitrary particle's motion to a combination of free-drift motion and centripetal motion through the coordinate system rotation and using a general linear interpolation tool developed at the Advanced Photon Source (APS). This method has been tested by tracking particles through conventional magnetic elements (dipole, sextupole, etc.) to verify reference coordinate system conversions, tracking accuracy, and long-term tracking stability. Results show a very good agreement between this new method and the traditional method. This method is not designed to replace mature traditional methods that have been used in most tracking codes. Rather, it is useful for magnets with complicated field profiles or for studying edge effects.