基于MLP神经网络的数控铣床几何误差补偿方法

Geometric Error Compensation for CNC Milling Machine Based on MLP Neural Network

针对三轴数控铣床加工工件的几何误差问题,提出一种基于多层感知器(MLP)神经网络的误差补偿方法。首先,设定铣床沿X轴和Y轴方向对工件进行铣削加工,通过3D坐标测量机测量刀头在Z轴上的定位误差数据。然后,利用这些数据来训练MLP神经网络模型,拟合出三维的误差曲面。最后,根据获得的误差曲面,对铣床加工时的刀头坐标进行实时校正,以此提高加工精度。实验结果表明,提出的方法能够对机床加工误差进行精确地补偿,具有有效性和可行性。

For the issue that the geometric error in 3-axis CNC milling machine, an error compensation method based on multilayer perceptron （MLP） neural network is proposed. First of all, it sets the machine milling workpiece in the X axis and Y axis direction, and uses the 3D coordinate measuring machine to measure the positioning error data in Z axis. Then, it uses these data to train MLP neural network model, so as to fit the three-dimensional error surface. Finally, the real-time correction of the cutting tool coordinate is processed according to the error surface obtained, so that to improve machining accuracy. The experimental results show that the proposed method can accurately compensate the machining error, and it is effective and feasible.