Estimation of Global Sizes of Cylinder Based on the Neural Network Regression

Abstract

In ISO 14405-1, the global sizes are defined, which include the least-squares, minimum circumscribed, maximum inscribed and minimax diameters for cylinder. The evaluation models of the global sizes were built based on the cylindrical coordinate measuring principle. The pseudo global sizes of cylinder can be obtained on the basis of the extracted data in the cylindricity measurement and a given value. The global sizes of cylinder can be precisely estimated by using the trained models of neural network regression between pseudo global sizes and global sizes, and the evaluation results showed that the estimated global sizes can meet their measurement accuracy requirements.