| Holes and shafts are widely used basic parts.With the rapid development of the manufacturing industry,the requirements for their sizes and geometric accuracies and their standardized drawings indicating are becoming higher and higher.The mating performance of the hole and the shaft is determined by their global sizes and calculated sizes formed by the actual sizes and form errors.Due to the lack of measuring instruments for the global sizes and calculated sizes of cylinders,the relationship between cylinder errors and sizes was studied based on the neural network regression analysis method in this thesis.Through the measurement of actual sizes and form errors of the cylinders,this will lay a foundation for obtaining their global sizes and calculated sizes indirectly.Based on the roundness extraction strategy,the simulation model of cylinder feature’s roundness profiles was established,and the simulation program of the cylinder feature’s profiles was developed by using Matlab software.According to the given simulation parameters,the cylinder feature’s profiles were simulated in batch and their actual sizes,global sizes,calculated sizes,roundness errors and cylindricity errors were evaluated.The maximum,minimum and mean values of their actual sizes,roundness errors,cylindricity errors,calculated sizes and global sizes were statistically analyzed.Based on BP neural network,the relationship model between actual sizes,roundness errors and circumference diameters or area diameters and the relationship model between actual sizes,cylindricity errors and global sizes or volume diameters were established,respectively.The corresponding programs for the relationship models were compiled by using Matlab software,and the models were trained on the basis of the evaluation results of sizes and form errors of the simulated cylinder features’ profiles.The training results showed that the model can meet the needs of the measurement of the global sizes and calculated sizes of cylinders in accuracy.The cylinder features’ roundness profiles of the designed and manufactured samples of holes and shafts were extracted by using the cylindricity measuring instrument,and the radial sizes from their corresponding profiles to the rotary axis were obtained for holes and shafts through calibration using the standard ring and the standard cylinder,respectively.Their singularities of the changed profiles were judged by the 3σ criterion and replaced by using the linear interpolation method.The actual sizes,global sizes,calculated sizes and form errors were evaluated,and the evaluation results were plugged into the established relationship models to verify their usability.Their verification results showed that the evaluation accuracies of the global sizes and calculated sizes based on established models can meet their indirect measurement with higher prediction.Therefore,the study of the relationship between the cylindrical form errors and the sizes based on the neural network may provide the technical support for the popularization and application of globalsizes and calculated sizes in mechanical manufacturing and the improvement of product quality. |