基于模糊策略的参数自整定迭代学习方法应用研究

发布时间：2019-06-10 02:16

【摘要】：当今时代IC已经成为全球电子信息行业的核心,而其中光刻机是关键设备之一。本文针对光刻机双工件台中直线电机重复周期性运行的特点及系统对光刻机性能的要求,对电机的精密控制问题进行了研究。提出了基于模糊策略的增益自整定迭代学习控制方法,且基于遗传算法对其进行了优化,并验证了算法的有效性。首先,分析了光刻机系统的整体情况以及双工件台系统的结构和功能,根据具体的扫描曝光过程流程及各电机在此过程中所承担的功能,设计了电机运行的三阶S曲线。此外,计算得到了本课题中所研究的永磁直线电机基于PARK变换和矢量控制的数学模型。其次,根据电机周期性运行的特点及性能要求,将迭代学习控制引入到双工件台直线电机的控制当中,经过数次迭代后可以明显提高精度,但收敛速度还有提升的余地。因此,本文提出了一种基于模糊策略的增益自调整迭代学习控制,即利用Mamdani型模糊控制器调整迭代学习的增益,该方法可以在不太影响精度的同时,大幅提高收敛速度。但该控制方法在仿真中出现了震荡的现象,其控制效果有很大的改进余地,模糊控制规则还可以进一步优化。针对上述问题,本文应用遗传算法优化模糊规则,建立模糊迭代学习控制方法。即利用遗传算法优化了模糊规则后件中的十个参数,经过若干代的进化之后,产生一组最优解,将该最优解对应的模糊规则用于Mamadani型模糊控制器,利用模糊控制器的输出调整迭代学习的增益,并进行仿真分析。仿真结果表明该方法稳定性很好,改善了专家经验型模糊迭代学习控制中出现的振荡现象。收敛速度也与专家经验的模糊迭代学习控制相近,且最终的控制精度也有一定的提高,说明了算法的有效性。此外,本文还对模糊迭代学习控制的抗干扰能力进行了仿真分析,优化后的模糊迭代学习控制在干扰下,也能取得良好的控制效果。最后,本文对X向直线电机设计了三组实验,普通的迭代学习控制和两组基于模糊策略的迭代学习控制。通过实验可以得出以下规律:普通的迭代学习控制可以在收敛速度很慢的条件下保证很高的精度,但收敛速度很难提高,而基于模糊策略的迭代学习控制可以大幅度地提高收敛速度。尤其是基于GA优化算法改进的模糊迭代学习控制方法,达到了收敛速度和控制精度完美的结合,找到了一个平衡点,可以应用于重复运行某一曲线的直线电机的控制上,例如本实验室光刻机双工件台的直线电机。可以收到满意的控制效果。
[Abstract]:Nowadays, IC has become the core of the global electronic information industry, and lithography machine is one of the key equipment. In this paper, the precision control of the linear motor is studied according to the characteristics of repeated periodic operation of the linear motor in the double workpiece platform of the lithography machine and the requirements of the system for the performance of the lithography machine. A gain self-tuning iterative learning control method based on fuzzy strategy is proposed and optimized based on genetic algorithm, and the effectiveness of the algorithm is verified. Firstly, the overall situation of the lithography machine system and the structure and function of the duplex system are analyzed. According to the specific scanning and exposure process flow and the functions of each motor in this process, the third-order S curve of the motor operation is designed. In addition, the mathematical model of permanent magnet linear motor based on PARK transform and vector control is obtained. Secondly, according to the characteristics and performance requirements of the periodic operation of the motor, the iterative learning control is introduced into the control of the linear motor with two workpieces. After several iterations, the accuracy can be improved obviously, but the convergence speed still has room to improve. Therefore, in this paper, a gain self-tuning iterative learning control based on fuzzy strategy is proposed, that is, Mamdani fuzzy controller is used to adjust the gain of iterative learning. This method can greatly improve the convergence speed without much affecting the accuracy. However, the control method has the phenomenon of concussion in the simulation, its control effect has a lot of room for improvement, and the fuzzy control rules can be further optimized. In order to solve the above problems, genetic algorithm is used to optimize fuzzy rules and fuzzy iterative learning control method is established in this paper. That is, the genetic algorithm is used to optimize the ten parameters in the latter part of the fuzzy rule. After several generations of evolution, a set of optimal solutions is generated, and the fuzzy rules corresponding to the optimal solution are applied to the Mamadani fuzzy controller. The output of fuzzy controller is used to adjust the gain of iterative learning, and the simulation analysis is carried out. The simulation results show that the method is stable and improves the oscillations in expert empirical fuzzy iterative learning control. The convergence rate is also similar to the fuzzy iterative learning control with expert experience, and the final control accuracy is also improved to a certain extent, which shows the effectiveness of the algorithm. In addition, the anti-interference ability of fuzzy iterative learning control is simulated and analyzed in this paper. The optimized fuzzy iterative learning control can also achieve good control effect under interference. Finally, three groups of experiments, ordinary iterative learning control and two groups of iterative learning control based on fuzzy strategy, are designed for X-direction linear motor. Through experiments, the following laws can be obtained: ordinary iterative learning control can ensure high accuracy under the condition of slow convergence speed, but the convergence speed is difficult to improve. The iterative learning control based on fuzzy strategy can greatly improve the convergence speed. In particular, the improved fuzzy iterative learning control method based on GA optimization algorithm achieves the perfect combination of convergence speed and control accuracy, and finds a balance point, which can be applied to the control of linear motor running a certain curve repeatedly. For example, the linear motor of the double workpiece table of the lithography machine in our laboratory. Satisfactory control effect can be obtained.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP273

【参考文献】