【摘要】 针对机床进给系统热误差,提出了一种复合预测方法。首先使用神经网络对丝杠热变形量进行建模,然后以多项式拟合法来求得平台轴向热误差y与丝杠实际热变形量x之间的关系,建立平台轴向热误差预测模型,称作"两步法"。两步法与BP神经网络直接预测平台轴向热误差的方法(直接法)分别对实验进行预测,将两方法预测结果与实测平台轴向热误差对比。结果表明,实验过程中行程发生变化时,直接法预测得到的热误差残差在-5.4～6.6μm间波动,两步法预测得到的热误差残差在-3.1～2.2μm内波动,两步法预测精度比直接法高了约126%,有较强的工程应用价值。更多还原

【Abstract】 A composite prediction method is proposed for the thermal error of the machine feed system. Firstly, the neural network is used to model the thermal deformation of the screw. Then the polynomial fitting method is used to find the relationship between the platform axial thermal error y and the actual thermal deformation x of the lead screw. Establish a platform axial thermal error prediction model called "two-step method". Two-step method, and BP neural network method for directly predicting axial thermal error of platform(direct method), respectively predicting experiment. And the prediction results of the two methods are compare with measured values of axial thermal error of the platform. The results show that when the stroke changes during the experiment, the thermal error residual predicted by the direct method fluctuates between -5.4 ～ 6.6 μm, and the thermal error residual predicted by the two-step method fluctuates within -3.1～2.2 μm. The prediction accuracy of the two-step method is about 126% higher than the direct method and has a strong engineering application value.更多还原