(A) study on alignment automation of multi-lens using neural network and interferometer : application to focusing unit in near field recording system = 신경회로망과 간섭계를 이용한 다군 렌즈 정렬 자동화에 관한 연구 : 근접장 정보저장 기기에의 응용application to focusing unit in near field recording system
In this thesis, a methodology of alignment automation for optical components - in details multi-lens system- is proposed. To apply the proposed methodology to an optical system, the near-field recording (NFR) system was chosen. The optical pick-up head - focusing unit (FU) of the NFR system is very small size. The components of the FU also are very small lenses. Due to the very small size of the FU, it is difficult to assemble well and the alignment among optical components is significant problem.
Therefore, high precision measurement (evaluation) and assembly are necessary to guarantee the optical performance of the FU. The auto-alignment methodology is described as following sequence
First, through the evaluation of the optical performance of manually assembled FU, the importance of assembly problem is emphasized again and then the error analysis for the FU was carried out. This error analysis is independent and composite error analysis and the trend and characteristic of each error is obtained.
Second, based on the error analysis results, the methodology of alignment automation using the neural network and interferometer is proposed. The evaluation system (interferometer) including the FU is modeled and the interference fringe patterns of the FU are analyzed according to errors. Using the obtained error trend, a proper feature extraction method of the fringe patterns is decided and a learning process using the neural network with multilayer perceptron and an error-back-propagation algorithm proceeded.
Third, using the above simulated results, we set up the experimental verification system and carried out the auto-alignment experiment. The system is based on the interferometer. For a position control of the FU, multi-axis pico-motor stages were used. The pico-motor has the position resolution less than 30 nm. For an exact closed-loop position control, we used two kind of capacitive type gap sensor and the sensor resolution was 50 nm and 10 nm each. Initially...