Machine learning entails a broad range of techniques that can provide benefits to the field of accelerator physics, including the beam optics controls. This talk will demonstrate a successful development of a new ML-based approach for optics control on the example of LHC optics measurements and corrections. These novel methods address such challenges as the detection of instrumentation faults, estimation of local magnet errors, measurements denoising and reconstruction of advanced optics observables. Along with the introduction to the relevant ML concepts, details on incorporating them into the optics control at the LHC and presentation of the achieved results, further possibilities of applying ML in beam dynamics studies will be discussed.
BE-ABP-LAF