Grant: Subatomic Physics Envelope - Group

This proposal aims to provide an integrated solution to problems often severely compromising the optimization processes in designing, building and operating an accelerator or beam delivery system: (1) Design tools assembled ad hoc and sub-optimal, with limited scope, consistency, and flexibility. Time and efforts wasted in single-use applications. (2) Absence of unifying software infrastructure to interconnect, integrate, cross-check, and archive separate programs modeling distinct physical processes. (3) Insufficient effort for truly global optimization accounting for sub-system interplay and trade-off among comprehensive set of design and operating parameters. (4) Commissioning and operational strategies not integrated at ground level, leading to surprises, squandered resources and costly retrofits. (5) Robustness of design not rigorously treated in design processes, much less the study of error recoverability, a measure of control/feedback system effectiveness. The proposed program provides a platform for optimizing global accelerator performance, independent of machine configuration specifics or modeling tools, with integrated operational scenarios, and rigorous treatment of errors and recoverability. The platform is structured to anticipate very general optimization configurations, for example allowing global optimization to transport beams with diverse characteristics under the same operating parameters, highly nontrivial by conventional methods. The main optimization engine is a customized genetic algorithm, with possible alternative algorithms. The program will be tested and run on WestGrid, the high performance computing grid across western Canada. Immediate applications include designing the TRIUMF Electron Linac (E-Linac) for simultaneous delivery of radioactive ion beam and free electron laser through energy recovery. This platform aims to advance the current paradigm of accelerator design by introducing true global optimization into the process, integrating ad hoc tools into unified user-friendly ones via high level software structure, with the prospect of evolving into automated accelerator global-optimum online control and spawning research on algorithm/architecture in large scale optimization.