Measurement and calibration method for an optical encoder based on adaptive differential evolution-Fourier neural networks

A high-precision measurement and calibration device is proposed in this paper. The resolution of this device can reach 18 binary bits, and it can measure whether or not optical encoders reach their nominal accuracy. The method based on the adaptive differential evolution-Fourier neural network (ADE-FNN) is proposed to improve the accuracy of optical encoders. This method makes full use of the FNN to establish an error compensation model for optical encoders and introduces an ADE algorithm to optimize the weights of the FNN. Compared to a nonlinear least-squares method, a back propagation neural network and a standard FNN, this method possesses many advantages, such as the fine nonlinear approximation capability, faster convergence speed and easiness of finding the global optimum. Experimental results demonstrate that after being calibrated by this method, significant improvement regarding the accuracy of optical encoders can be achieved.