轧辊辊型曲线的评定方法

Evaluation of the Profile Curve of Mill Roll

提出把轧辊辊型测量曲线评定转换为对形状曲线w(x)的构造和噪声曲线y(x)的评定这两个过程。其中,形状曲线w(x)作为噪声曲线y(x)评定的基准曲线,可采用递归神经网络(GRNN)来进行构造。文中不但分析了此法的优点,还指出用递归神经网络能够捕捉局部高点。噪声曲线y(x)的评定采用的是基于控制线旋转的最小区域法。最后通过仿真试验证明了整个过程的可行性。

A new method to evaluate the profile curve of mill roll is described. The whole process is divided into two steps: one is to build shape curve w(x), the other is to evaluate the noise curve y(x). w(x) is used as the base curve to evaluate y(x), it is built by the general regression neural network (GRNN). The merits of the approach are identified and the GRNN can be used to catch local peak on the surface of mill roll is demonstrated. The noise curve y(x) is evaluated by the control line based on the extreme fit method. The final simulation proves the effectiveness of the whole process.