Morita, Y. Fukuda, M. Kanda, H. Washio, T. Yorita, T. JACoW Publishing Geneva, Switzerland 2673-5482 978-3-95450-212-7 10.18429/JACoW-CYCLOTRONS2022-THPO019 English 347-350 experiment ion-source brightness LEBT ECR Contribution to a conference proceedings 2023 2023-10 https://doi.org/10.18429/JACoW-CYCLOTRONS2022-THPO019 https://jacow.org/cyclotrons2022/papers/thpo019.pdf An enormous number of parameters are tuned during accelerator operation. The tuning is ultimately dependent on the operator’s knowledge and experience. Therefore, there is a risk that tuning time and accuracy may vary depending on the operator. This tuning difficulty is an extremely important issue when implementing accelerometers in society, such as in medical applications. In this study, we developed an automatic tuning method using Bayesian optimization, one of the machine learning technique. The aim is to realize a tuning method that can supply beams in a short time with good reproducibility and comparable to manual tuning.