【作者】 谭文斌；

【导师】 李醒飞；

【作者基本信息】 天津大学 ， 测试计量技术及仪器， 2012， 博士

【摘要】 伺服系统中存在非线性摩擦,而温度变化不仅影响摩擦,同时还会导致热变形误差的产生,这些均对提高伺服系统的控制和定位精度有严重的影响。随着伺服系统的应用越来越广,对其精度要求也越来越高,减小甚至消除摩擦和温度变化对伺服系统的影响,已成为实现高精度伺服系统所必须解决的关键问题。本文针对这一问题,通过理论分析和实验分析相结合的方法,分别对体现温度影响的摩擦以及丝杠轴向热变形的建模和补偿等方面展开了深入研究。首先,完成了体现温度影响的摩擦模型的建模和辨识。采用一种基于稳态误差反推的方法分离并获取伺服系统中存在的转矩纹波和摩擦力矩;利用频谱分析、最小二乘等方法辨识转矩纹波,并通过运动控制器实现其补偿;利用遗传算法实现传统LuGre模型动静态参数的精确辨识;根据实验修正传统LuGre模型中的黏性摩擦,使其在高速阶段也能精确反映伺服系统的摩擦;研究温度变化对摩擦的影响,提出一种利用神经网络对LuGre摩擦模型的静态参数进行建模的方法,从而将温度直接引入摩擦模型,建立了一种体现温度影响的摩擦模型。然后,基于该修正模型对伺服系统的摩擦补偿策略展开了研究。一方面,采用前馈PD固定摩擦补偿的控制方法,提高了伺服系统的控制精度;另一方面,针对伺服系统中摩擦模型的参数变化,以及不确定非线性建模误差和其他扰动,提出了一种基于反演设计的自适应反演滑模的摩擦补偿策略,给出了其补偿控制器的设计方法,并证明了其渐进稳定性,实验证明该方法具有较强的适应性和鲁棒性,有效改善了伺服系统的跟踪精度。最后,对温度变化引起的丝杠轴向热变形误差进行分析和建模。补偿了螺距误差和反向间隙等几何误差;分析滚珠丝杠副的热特性,得出其热变形不仅由当时其内部热源的状态决定,还受到其之前状态的影响;提出一种采用自回归小波神经网络(SRWNN)进行丝杠轴向热变形误差建模的方法,获得了较好的建模效果,并通过实验验证了模型的有效性。更多还原

【Abstract】 Nonlinear Friction exists in the servo system. Temperature changes not only affect friction, but also cause to produce thermal deformation error, thus seriously affect control and positioning accuracy of servo system. With wider use of servo system, it requires much higher accuracy. Reduction or elimination errors in the servo system caused by changes in friction and temperature have become a critical issue in the realization of servo system. In this paper, aiming at this problem, combined theoretical analysis with experimental analysis, friction that reflects influence of temperature and modeling and compensation of axial thermal deformation of screw are deeply researched separately.First, modeling and identification of friction model that reflect temperature changes are accomplished. Backstepping method based on state error is used to separate and obtain torque ripple and friction moment in the servo system; spectral analysis, least squares and other methods are adopted in the identification of torque ripple, and motion controller is used in the compensation; genetic algorithm is used in the precise identification of dynamic and static parameters in the custom LuGre model; viscous friction in the custom LuGre model is corrected according to experiments, making friction in the servo system is accurately reflected even in high speed; influence on the friction by temperature changes is studied, and based on temperature observation on the key point, method of applying neural network to the modeling of static parameter of LuGre friction model is proposed, thus introducing temperature to the friction model, and friction model that reflects temperature changes is built.Then, friction compensation strategy of servo system based on the error correction model is researched. On the one hand, control method of feedforward PD constant friction compensation is adopted, and control accuracy of servo system is improved; on the other hand, aiming at parameter changes of friction model in the servo system, uncertain nonlinear modeling error and other disturbances, adaptive backstepping sliding mode friction compensation method based on the backstepping design is proposed, design method of compensation controllers is provided, and asymptotic stability is verified. Adaptability and robustness of this method is proved by the tests, effectively improve tracking accuracy of servo system. Last, axial thermal deformation error of screw caused by temperature variation is analyzed and modeled. Geometric errors such as pitch error and backlash are compensated;thermal characteristics of ball screw assembly is analyzed, and come to a conclusion that thermal deformation not only determined by the state of internal heat source at that time, but also affected by the previous state; method that use self-recurrent wavelet neural network (SRWNN) is presented to model axial thermal deformation error of screw, acquiring good modeling effect, and effectiveness of the model is verified by experiments.更多还原