Reconstructing Space-Charge Distorted IPM Profiles with Machine Learning Algorithms

Measurements of undistorted transverse profiles via Ionization Profile Monitors (IPMs) may pose a great challenge for high brightness or high energy beams due to interaction of ionized electrons or ions with the electromagnetic field of the beam. This contribution presents application of various machine learning algorithms to the problem of reconstructing the actual beam profile from measured profiles that are distorted by beam space-charge interaction. (Generalized) linear regression, artificial neural network and support vector machine algorithms are trained with simulation data, obtained from the Virtual-IPM simulation tool, in order to learn the relation between distorted profiles and original beam dimension. The performance of different algorithms is assessed and the obtained results are very promising for testing with simulation data.