IPAC2015 - List of Authors (van der Geer, S.B.)

Paper	Title	Page
MOPJE075	Tracking Through Analytic Quadrupole Fringe Fields With GPT	489
	S.B. van der Geer, M.J. de Loos Pulsar Physics, Eindhoven, The Netherlands B.D. Muratori STFC/DL/ASTeC, Daresbury, Warrington, Cheshire, United Kingdom
	In the early design stages it is customary to work with a highly simplified analytic model to describe the beam line. Dipoles and quadrupoles are often based on hard-edged approximations. This is not only unrealistic, it also significantly slows down time-domain spacecharge tracking codes such as the General Particle Tracer (GPT) code. The underlying reason for the poor performance is that despite the fact that the simple hard-edged field equations are fast to evaluate, they force the integration process to use excessively small step sizes near the fields discontinuities in order to achieve the desired accuracy. In other worlds, the apparently simple equations turn out to be the most difficult ones to evaluate numerically. An obvious solution is to switch to field-maps, but this is not practical in the early design stages. In this contribution we show a new solution implemented in the GPT code based on analytical expressions for the fringes where the transverse size of the magnet is properly taken into account. In addition to producing more realistic results, the smooth fields increase tracking speed by over an order of magnitude for typical test cases.
DOI •	reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2015-MOPJE075
Export •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

MOPJE076	Multi-objective Genetic Optimization with the General Particle Tracer (GPT) Code	492
	S.B. van der Geer, M.J. de Loos Pulsar Physics, Eindhoven, The Netherlands
	In a typical design process there are a large number of variables, external constraints, and multiple conflicting objectives. Examples of the latter are short pulse, high charge, low emittance and low price. The classical solution to handle such problems is to combine all objectives into one merit function. This however implicitly assumes that the tradeoffs between all objectives are a-priori known. Especially in the early design stages this is hardly ever the case. A popular solution to this problem is to switch to multi-objective genetic optimization algorithms. This class of algorithms solves the problem by genetically optimising an entire population of sample solutions. Selection and recombination operators are defined such that the output, the so-called Pareto front, only includes solutions that are fully optimized where no objective can be improved without degrading any other. Here we present numerical studies and practical test runs of the genetic optimizer built into the General Particle Tracer (GPT) code.
DOI •	reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2015-MOPJE076
Export •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

Paper

Title

Page

Tracking Through Analytic Quadrupole Fringe Fields With GPT

489

S.B. van der Geer, M.J. de Loos
Pulsar Physics, Eindhoven, The Netherlands
B.D. Muratori
STFC/DL/ASTeC, Daresbury, Warrington, Cheshire, United Kingdom

In the early design stages it is customary to work with a highly simplified analytic model to describe the beam line. Dipoles and quadrupoles are often based on hard-edged approximations. This is not only unrealistic, it also significantly slows down time-domain spacecharge tracking codes such as the General Particle Tracer (GPT) code. The underlying reason for the poor performance is that despite the fact that the simple hard-edged field equations are fast to evaluate, they force the integration process to use excessively small step sizes near the fields discontinuities in order to achieve the desired accuracy. In other worlds, the apparently simple equations turn out to be the most difficult ones to evaluate numerically. An obvious solution is to switch to field-maps, but this is not practical in the early design stages. In this contribution we show a new solution implemented in the GPT code based on analytical expressions for the fringes where the transverse size of the magnet is properly taken into account. In addition to producing more realistic results, the smooth fields increase tracking speed by over an order of magnitude for typical test cases.

DOI •

reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2015-MOPJE075

Export •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

MOPJE076

Multi-objective Genetic Optimization with the General Particle Tracer (GPT) Code

492

S.B. van der Geer, M.J. de Loos
Pulsar Physics, Eindhoven, The Netherlands

In a typical design process there are a large number of variables, external constraints, and multiple conflicting objectives. Examples of the latter are short pulse, high charge, low emittance and low price. The classical solution to handle such problems is to combine all objectives into one merit function. This however implicitly assumes that the tradeoffs between all objectives are a-priori known. Especially in the early design stages this is hardly ever the case. A popular solution to this problem is to switch to multi-objective genetic optimization algorithms. This class of algorithms solves the problem by genetically optimising an entire population of sample solutions. Selection and recombination operators are defined such that the output, the so-called Pareto front, only includes solutions that are fully optimized where no objective can be improved without degrading any other. Here we present numerical studies and practical test runs of the genetic optimizer built into the General Particle Tracer (GPT) code.

DOI •

reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2015-MOPJE076

Export •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)