Detection of Surface Roughness of Nickel-Plated Copper Coil by SA-BP Neural Network
In order to effectively detect the surface roughness of the nickel plating rectangular copper coils,by industrial camera,microscope lens,point light source equipment such as composition of the surface of the hardware system for coil image,visual in-spection method based on gray level co-occurrence matrix,based on image processing techniques to extract eight based on gray level co-occurrence matrix feature parameters,combined with the actual roughness values to establish the experimental data-base,analyses the change rules of characteristic parameters and the actual roughness value;Abstract:BP neural network is easy to make weights and thresholds fall into local optimal solution,resulting in inaccurate detection results,and so on.SA algorithm is used to optimize the initial weights and thresholds of BP neural network,and a detection model of SA-BP neural network is built.According to the training results,the training MSE decreased from 0.000139 of BP model to 0.000023,and the number of iterations decreased by 22,indicating that SA-BP model has faster convergence speed and better network model.According to the detection results,the maximum detection error range decreased from 0.21μm of BP model to 0.13μm,and the relative error mean decreased from 5.41% to 3.45%,indicating that SA-BP model has higher detection stability and accuracy.
NETL
NSTL
万方数据
维普
