An accurate model for active surface system of milliter/submilliter wave radio telescope

The active surface system plays a crucial role in the high-precision observation of large aperture millimeter/submillimeter wave radio telescopes, as environmental factors can lead to deformation of the primary reflector surface (PRS). The rapid and accurate measurement and correction of PRS deformation is the key to active surface systems. To enhance both the speed and accuracy of PRS deformation measurement, this paper proposes to construct a machine learning model based on random forest regression analysis. Using photogrammetric techniques, a large amount of actuator disturbance data and PRS point position data are obtained and trained. The effectiveness of the model for measuring PRS deformation is verified through comparative experiments. The results demonstrate that when six actuators apply 100 μm level disturbances to PRS single panel, the model predicts its deformation (RMS) with an error of less than 20 μm relative to the photogrammetric measurements. In addition, the model can also evaluate the efficiency of the active surface system in correcting PRS deformation through actuators.