基于神经网络的快速成型系统速度控制

Velocity Control of RPM System Based on ANN

提出了一种基于神经网络的数据融合方法 ,用于提高快速成型系统的扫描速度、改进系统加工的轮廓精度与系统运行的平稳性 .该方法中神经网络的学习样本来源于两个方面 :一方面是利用现有系统的进给控制算法 ,它主要考虑了系统沿曲线路径加工时的速度控制规则 ;另一方面 ,采用可变加速度策略从理论上设计直线运动的进给速度 .该方法对路径特征的描述具有位置不变性和旋转不变性的特点 .对该算法的数字仿真结果表明 :与常规的控制算法相比较 ,融合了两类样本的神经网络算法能更有效地提高系统运行的速度 。

BP(Back Propagation) neural network based on data combined technique for velocity control of a RPM(Rapid Prototyping & Manufacturing) system is considered in this paper. In order to reduce the computation time and complexity greatly, a group of neural nets is employed for different velocity ranges respectively. The learning sample used to train the nets is acquired from two sources. One part is sampled from a practical system, whose velocity generated in the kinematics principle is relatively low along with curve path. The other part is calculated for the laser head of the bi axis system moving from one point to another along with straight line segments by the formula, which is derived from kinematics theory. A variable acceleration method is introduced to obtain high velocity within short period without shocking the system simulation on the nets combined with the two different sources is performed and the result report is also described.