A high-precision and efficient adaptive alignment method for Cassegrain dual mirror optical system based on machine learning algorithms

Abstract

The Cassegrain telescope is widely used in aerospace exploration equipment, characterized by compact structure, complex optical path, and high imaging quality. However, due to the difficulty in establishing an accurate correspondence between the relative pose and imaging quality of a multi mirror group with real machining errors, the current Cassegrain telescope assembly process is very difficult, with blind operation, time-consuming, and low accuracy. This article proposes a high-precision adaptive alignment method for Cassegrain dual mirror optical systems based on machine learning algorisms, and an experimental adaptive alignment system with uncoupled degrees of freedom is developed. Firstly, the overall architecture of the adaptive alignment method is proposed consists of detection, calculation and alignment modules. In the detection module, the Zernike polynomial coefficient of wavefront aberration of the optical system is detected online through the interferometer, meanwhile the pose coordinates of the secondary mirror is accurately fed back through a 6-DOF nanoscale micro mechanism. In the calculation module, machine learning algorithm is applied to build a nonlinear mapping model between the Zernike coefficient and the pose coordinates of the secondary mirror. In the alignment module, the pose coordinates of the secondary mirror can be forced to adjust to the target position. Then, during the real alignment process, the Zernike coefficient test data of the optical system alignment process is monitored in real time, and the nonlinear mapping model is employed to calculate the pose coordinates and then the misalignment of the secondary mirror. Finally, the alignment module is driven to execute the pose correction according to the calculated misalignment value, realizing a high-precision adjustment of the Cassegrain dual mirror optical system. Experimental results shows that the average alignment time cost can be dramatically reduced from around 7 days using the current manual alignment method to just 30 minutes using the proposed adaptive alignment method for achieving a current standard alignment accuracy of wavefront aberration root mean square (RMS) less than 0.1λ, which greatly improves the assembly efficiency. This study proposes a new method for high-precision and efficient alignment of optical systems based on artificial intelligence and contributes to the efficiency improvement for optical assembly.