论文部分内容阅读
分析了CMAC神经网络应力盘智能控制中面形表征及检测的特点,提出了以Zernike多项式系数用于表征应力盘盘面面形的方法,以微位移阵列传感器检测得到的原始数据经插值拟合重构出盘面面型数据,再经Gram-Schimdt正交化后拟合出的Zernike多项式的系数,以此作为CMAC神经网络的输入样本来完成CMAC神经网络的训练。并在实验中初步验证了此方法的可行性。
The characteristics of surface shape representation and detection in intelligent control of stress plate of CMAC neural network are analyzed. A method of using Zernike polynomial coefficients to characterize the surface shape of stress plate is proposed. The original data detected by micro-displacement array sensor is interpolated by interpolation The surface area data of disk is constructed, then the coefficients of Zernike polynomials fitted by Gram-Schimdt orthogonalization are used as the input samples of CMAC neural network to complete the training of CMAC neural network. In the experiment, the feasibility of this method is verified.