The Joint Accelerator Conferences Website (JACoW) is an international collaboration that publishes the proceedings of accelerator conferences held around the world.
@inproceedings{donegani:ibic2021-tupp37,
author = {E.M. Donegani},
title = {{Machine-Learning Based Temperature Prediction for Beam-Interceptive Devices in the ESS Linac}},
booktitle = {Proc. IBIC'21},
pages = {306--308},
eid = {TUPP37},
language = {english},
keywords = {proton, linac, diagnostics, simulation, database},
venue = {Pohang, Rep. of Korea},
series = {International Beam Instrumentation Conference},
number = {10},
publisher = {JACoW Publishing, Geneva, Switzerland},
month = {10},
year = {2021},
issn = {2673-5350},
isbn = {978-3-95450-230-1},
doi = {10.18429/JACoW-IBIC2021-TUPP37},
url = {https://jacow.org/ibic2021/papers/tupp37.pdf},
note = {https://doi.org/10.18429/JACoW-IBIC2021-TUPP37},
abstract = {{’Where there is great power [density], there is great responsibility*.’ The concept holds true especially for beam-intercepting devices for the ESS linac commissioning. In particular, beam-intercepting devices will be subject to challenging beam power densities, stemming from proton energies up to 2 GeV, beam currents up to 62.5 mA, pulses up to few milliseconds long, and repetition rates up to 14 Hz. Dedicated Monte Carlo simulations and thermo-mechanical calculations are necessarily part of the design workflow, but they are too time-consuming when in need of rapid estimates of temperature trends. In this contribution, the usefulness of a Recurrent Neural Network (RNN) was explored in order to forecast (in few minutes) the bulk temperature of beam-interceptive devices. The RNN was trained with the already existing database of MCNPX/ANSYS results from design studies. The feasibility of the method will be exemplified in the case of the Insertable Beam Stop within the Spoke section of the ESS linac.}},
}