摘要
动力学因素,如摩擦力和惯性力,在叠层实体成型加工中常影响激光机床的切割精度。提出的动力学补偿方法结合了闭环位置控制和计算力矩控制两者的优点。既可以避免对名义轨迹的偏差,又可以补偿动力学因素对精度的影响。在每个原动件控制中,用附加的速度前馈来实现动力学补偿。多层前馈型神经网络用来实现机构的逆动力学模型。用周期函数,有限项傅立叶级数,作为激励函数来获取训练样本。复杂的动力学参数辨识过程成为神经网络权值的监督学习过程。实验结果表明,本文提出的方法对提高激光切割的轨迹精度和切口角度精度是有效的。
Some dynamic factors, such as friction and inertia forces, affect the cutting accuracy of laser machine in laminated solid manufacture. The proposed dynamic compensation method combines the advantages of closed-loop position control and computed torque control. The deviation from nominal trajectory is avoided and the effects of dynamic factors on cutting accuracy are compensated. The dynamic compensation is realized by velocity feed-forward for each actuator. Multilayer feed-forward neural network is employed to realize the inverse dynamics model. A periodic function, finite Fourier series, is used to activate the actuator for obtaining training samples. The complicated dynamic parameter identification process now becomes the learning process of neural network connecting weights under supervision. Experimental results have shown that the method is valid for improving the trajectory accuracy and tangent angle accuracy.
出处
《机械科学与技术》
CSCD
北大核心
2005年第4期484-487,共4页
Mechanical Science and Technology for Aerospace Engineering
关键词
叠层实体成型
激光切割
动力学补偿
神经网络
Laminated solid formation
Laser cutting
Dynamic compensation
Neural network