【作者】 周伦才；

【导师】 邬再新；

【作者基本信息】 兰州理工大学 ， 机械制造及其自动化， 2008， 硕士

【摘要】 采用数控加工的目的是提高产品加工的精度和效率,因此加工精度是数控机床最重要的性能指标之一,而对机床实施误差补偿是提高机床加工精度较为有效的方法,但由于数控加工过程具有复杂性、非线性、不确定性等特点,用传统的基于机床成形系统精确数学模型的方法已经难以获得良好的误差补偿效果。本文就是以如何提高数控机床误差补偿精度为目的而展开的。以国内外研究为基础,应用多体系统误差分析理论建立了数控机床理想运动模型、误差情况下的运动模型、刀具空间姿态理想运动模型以及刀具空间姿态运动误差模型。并以DMG公司生产的五轴数控万能镗铣床DMU 70V(TPPPBRRW)为例给出了成形点空间运动误差模型和刀具姿态运动误差模型。机床误差参数的正确辨识是误差补偿的前提条件之一,本文利用多体系统运动学理论对平动轴和转动轴的几何误差进行了正确辨识。为了提高误差补偿效果,本文在分析神经网络学习机理的基础上,利用神经网络良好的逼近能力、泛化能力及自学习能力的特点,通过对数控系统进行神经网络辨识,对误差补偿技术和误差控制的神经网络实现方法进行分析,建立了神经网络误差补偿模型。结合了双频激光干涉仪位移测量和直线度测量及三点法回转误差测量法,综合基于多体系统运动学理论建立的误差模型,以及机床几何误差的辨识,利用Malab软件对测量参数进行处理,获得了较为全面的网络训练样本,进一步提高了网络的精度。通过仿真试验验证了机床成形点空间位置误差模型的正确性和神经网络误差补偿的可行性,对补偿前后的结果分析可以看出,将神经网络技术应用在数控机床误差补偿控制中是可行的,与传统误差补偿方法相比,基于神经网络的数控机床误差补偿具有补偿精度较高、稳定性较好的特点。更多还原

【Abstract】 The purpose of adopting NC machining is to improve precision and efficiency , so the machining accuracy is one of the most important performance indexes for NC machine. Error compensation is a good way to improve accuracy of NC machining. But the process of NC machining is complicated, non-linear and uncertain. It almost can’t be described precisely by mathematical model so that a well result of error compensation using traditional method cann’t be got. Just for the purpose of enhancing error compensation accuracy of NC machine tools, this thesis was carried out.Based on the domestic and foreign research about error compensation, the ideal kinematic model and nonideal kinematic model for NC machine tools, the ideal gesture kinematic model and gesture error kinematic model for cutting tool were established by using the method of movement error analysis of multi-body system (MBS ).The above models were given for five-axes NC boring-milling machine DMU 70V (TPPPBRRW) manufactured by DMG. Correctly identifying of geometric error parameters is one of the most important preconditions for error compensation. The geometric error parameters were identified of translatory axes and rotary axes by using of MBS.To improve precision of error compensation, a model of error compensation based on Artificial Neural Networks (ANN) theory and identified with ANN through NC system was built with ANN, which possesses excellent approximation ability, generalization ability and self-learning ability. By combining dual-frequency laser interferometer used to measure displacement and straightness, three points method of measure rotary error in which are brought error calculation models based on MBS, identification of machine geometrical error and processing for measure parameters with Matlab software, the comprehensive training samples was gained, which further raised the network precision.The simulation demonstrates the correctness of volumetric kinematic error models for forming point and the feasibility of the ANN error compensation. Analyzing the result of the error compensation, conclusion that technology of ANN can be applied in controlling of NC machining error compensation can be gotten. Compared with traditional method of error compensation, the method of NC machining error compensation based on ANN possesses characteristics of high precision and good stability.更多还原