摘要
在应用激光技术加工复杂曲面时,通常以采样点集为插值点来建立曲面函数,然后实现曲面上任意坐标点的精确定位。人工神经网络的BP算法能实现函数插值,但计算精度偏低,往往达不到插值精确要求,造成较大的加工误差。提出人工神经网络的共轭梯度最优化插值新算法,并通过实例仿真,证明了这种曲面精确定位方法的可行性,从而为激光加工的三维精确定位提供了一种良好解决方案。这种方法己经应用在实际中。
In laser machining on complex curved surface, interpolation sets of sampling coordinate measured points are used to construct the curved surface function, by which arbitrary coordinate of curved surface can be determined. BP algorithm about artificial neural network may be employed to carry out interpolation, but it is hard to reach machining precision due to low calculating precision leading to high orientation error. A novel interpolation method, based on conjugate gradient optimization algorithm about artificial neural network, is proposed and introduced, and feasibility of the method is verified via orientation simulation of the curved surface. The new algorithm will be used to provide a solution of the accurate three-dimensional orientation for laser machining.
出处
《机械工程学报》
EI
CAS
CSCD
北大核心
2002年第11期143-146,共4页
Journal of Mechanical Engineering
基金
中国科学院创新工程重大项目(KGCX1-11)
辽宁省博士起动基金(2001102017)资助项目。
关键词
激光加工
定位
共轭梯度优化算法
曲面仿真
神经网络
函数逼近
Conjugate gradient optimization algorithm Curved surface simulation Neural network Function approximation
