An Active Flexure Compensation Method for LAMOST spectrograph based on BP-Neural Network

Abstract

The Large sky AreaMulti-Object fiber Spectroscopic Telescope (LAMOST) is a Chinese national scientific research facility operated by National Astronomical Observatories, Chinese Academy of Sciences (NAOC). The LAMOST survey for the Milky Way Galaxy and extra-galactic objects has been carried out for several years. The accuracies in measuring radial velocity are expected to be 5 km s−1 for the low resolution spectroscopic survey (R = 1800), and 1 km s−1 for the medium resolution mode. The stability of spectrograph is the main factor affecting the accuracies in measuring radial velocity, so an Active Flexure Compensation Method (AFCM) based on Back Propagation Neural Network (BPNN) is proposed in this paper. It utilizes a deep BP (4-layer, 5-layer etc.) model of thermal-induced flexure to periodically predict and apply flexure corrections by commanding the corresponding tilt and tip motions to the camera. The spectrograph camera system is adjusted so that the positions of these spots match those in a reference image. The simulated calibration of this compensation method analytically illustrates its performance on LAMOST spectrograph.