【作者】 沈浩；

【导师】 强文义； 王昕；

【作者基本信息】 哈尔滨工业大学 ， 控制科学与工程， 2010， 博士

【摘要】 蝶形半导体激光器是光通讯行业重要的元器件,是最常用的长距离传输光信号放大装置。但是一直以来蝶形半导体激光器的封装成本高居不下,极大地限制了光电器件行业的发展速度。限制蝶形半导体激光器封装自动化的瓶颈主要有两大技术:光纤耦合(光纤对准)技术和尾纤固定技术。光纤耦合的瓶颈在于,实际对准大多数由人工根据激光器输出功率的大小来判断光纤是否对准,并利用半自动或手动的调节平台进行微调。这样势必导致激光器封装速度慢,成品率低,封装成本居高不下。自动化的光纤耦合过程首先需从光功率计中提取功率误差信号,配合先进的光纤对准算法将其转换为光纤末端位置误差信号,再通过控制器驱动亚微米级精度的光学对准平台将光纤逐步调整至最佳耦合位置。尾纤固定一般采用激光焊接的方式,其难点在于焊后变形导致最佳对准位置再度偏离,借助于光纤对准平台的微调和智能的焊后偏移补偿算法,焊后偏移完全可以被自动地补偿至最小。本课题旨在从四个方面展开研究工作以解决光纤对准与尾纤固定这两大难题。第一个方面主要从光学对准平台上入手。当前的多自由度并联光学对准平台一般采用单轴独立控制,忽略了各轴之间的耦合误差,这导致光纤耦合不准,输出功率下降,同时对准速度也会降低。本课题对3自由度6支链PSS构型的并联微动平台进行了研究,通过建立全闭环位置反馈系统来消除轴间耦合误差。首先本课题运用经典的牛顿-欧拉法对6PSS构型的并联微动平台进行了动力学建模;根据本平台正交无姿态变化的特点对模型进行了简化并求出其位置正解与工作空间;借助于简化的动力学模型设计了滑模控制器,通过闭环反馈实现3自由度并联微动平台闭环系统的仿真,仿真结果显示滑模控制可以消除轴间耦合误差,轨迹跟踪性能良好。由于滑模控制存在明显的抖振现象,并且其鲁棒稳定的条件较为苛刻,另外考虑到系统在实际应用时一般采用计算机控制,因此设计一个既保证鲁棒性有可有效缓解抖振现象的离散滑模控制器就显得尤为必要。本课题在离散滑模控制的基础上提出了基于滑模预测思想的控制系统设计新思路。以不确定系统的名义模型作为滑动模态预测模型,利用当前及过去时刻的滑模信息预测将来时刻不确定因素对系统的影响,从而实现了滚动优化求解,对滑模控制进行实时校正,既显著削弱了抖振现象,又使得闭环控制系统具有很强的鲁棒性。仿真结果也证实了滑模预测控制的优越性。现有的光纤耦合算法大都不考虑角度耦合。由于角度调节平台的旋转中心与光纤末端不重合导致角度调整会引入线性方向的额外偏差,因此大大影响光纤耦合的速度。针对这个问题,本课题采用了相角跟踪角度耦合法,实时地补偿角度耦合造成的线性方向额外偏差。相角跟踪是一种可以在较小的搜索范围内迅速定位最佳耦合位置的对准算法,一旦光纤产生偏离,光功率计就会反馈输出功率偏差,此时算法可以迅速根据功率偏差信号提取位置偏差信号,并实时将偏离的误差补偿回去。实验结果的比较也进一步证明相角跟踪在角度耦合应用上的优势。焊后偏移难以消除,但是可以通过预先调节某些焊接参数使之被优化至最小。本课题通过大量的实验测量出:焊前预偏移、焊点位置、激光脉冲能量以及焊接初始位置这四个可调参数对焊后偏移的影响情况,设计了基于BP神经网络的焊后偏移智能预补偿系统。大量实验数据训练出的BP网络可以预测出相应参数下的焊后偏移大小,因此系统可以在焊接前用光学对准平台将待焊接光纤调整至预偏移位置,焊接后由于焊后偏移和预偏移相互抵消,光纤则被补偿至最佳耦合位置。实验证明通过神经网络的预补偿策略焊后偏移可以被补偿至0.4μm以内。更多还原

【Abstract】 As the most common optical signal amplifier, butterfly laser diode module is one of the important components in photo-communication. But the expensive packaging cost is always the bottleneck, and extremely limits the development of photo-electricity industry. The main reason is the manual or half-automatic packaging process. To improve the automatic packaging level, two main technologies should be considered. One is the fiber coupling (fiber alignment) technology. The other is fiber fixing technology. To improve the fiber coupling technology, a multi-axis optical alignment stage with highly positioning ability should be equipped. Also a quick and stable fiber coupling algorithm should be applied. The obstacle of fiber fixing technology is post welding shift (PWS). Research works should be focused on how to reduce the PWS or how to compensate the PWS.This thesis deals with the fiber coupling and fiber fixing technology mainly based on four different sections. First researching aim is optical alignment stage. In now days many multi-axis optical alignment stage only have a closed loop for single axis but not for the whole stage. This kind of control strategy could not take the axis coupling error into account. And the fiber alignment time would be increased or the fiber output power would be decreased. In this thesis a whole closed loop position feedback system is designed for a 3-axies parallel micro-stage to compensate the axis coupling error influence. A dynamic model of the 6-PSS parallel micro-stage is also discussed. Also the working space and forward solution are discussed based on kinematic analysis. At last simulation results have proved that with a sliding mode controller the axis coupling error can be deleted and the position tracking ability of the micro-stage is well.Conventional sliding mode control has its two inherent drawbacks. The first one is chattering and the second one is the strict robustness boundary condition. So a improved discrete sliding mode controller should be designed for a computer control system. And the discrete sliding model predictive algorithm overcomes the shortcomings of the conventional sliding mode controller. By predicting the future sliding mode function value and combining feedback correction and receding horizon optimization approaches, a discrete-time sliding mode control law for tracking problem is constructed. With the designed control law, closed-loop systems have strong robustness to matched or unmatched uncertainties as they eliminate chattering. Numerical simulations illustrate the validity of the proposed algorithm.Currently most fiber alignment algorithms are only suitable for lateral axes, and for angular alignment. Manual coupling is still widely used in product line. In recent years, engineers have tried to apply some automatic algorithms such as Hill-climbing algorithm to realize angular alignment. But since the pivot point is not always the end point of the fiber tip, an angular movement will cause an extra offset in lateral axis. Consequently, the whole alignment process will cost much more time. In this thesis, a novel automatic angular alignment method based on phase angle tracking is presented. Phase angle tracking is used to detect the optical power increasing direction and drive the collimator to the maximum power point. Experimental results show that the novel automatic angular alignment method can reduce the total time less than 160 seconds. It is obvious that this method is rapidly improved the efficiency in optical collimators assembly.Post welding shift can not be fully eliminated. But by optimizing some welding parameters, the PWS can be compensated or minimized. Based on enough welding experiments and the analysis by FEM, influence factors of PWS are studied, such as pre-welding shift, welding spot position, welding laser energy and initial position respectively. A BP neural network is constructed for analyzing efficiently the welding data, so that PWS can be predicted before welding. And then the fiber could be moved to an anti-direction shift of the predictive PWS. After welding the anti-direction shift will compensate the PWS. The experiment can validate that the PWS can be compensate into 0.4μm更多还原