Surface roughness prediction in robotic belt grinding based on the undeformed chip thickness model and GRNN method

As an important evaluation index of surface quality, surface roughness can directly affect the service performance of products, which makes it an ever-increasing concern in industries and academia nowadays. In robotic belt grinding, due to the characteristics of elastic contact between workpiece and contact wheel together with the random distribution of abrasive grains, it is of great difficulty to predict surface roughness accurately. Starting from the formation mechanism of surface roughness and combining with the intelligent algorithm, a surface roughness prediction model based on the undeformed chip thickness (UCT) and generalized regression neural network (GRNN) is proposed. Firstly, fully considering the flexible contact characteristic in belt grinding, the grinding depth is calculated based on the modified Preston equation, and furthermore, the UCT is obtained; then, UCT formula is decomposed according to the computability of its variables, and then, the modified UCT, average size of abrasive grains, and the normal grinding force are selected as the input parameters; finally, based on GRNN, the surface roughness prediction model is presented. The experimental results indicate a good agreement between the predicted and experimental values which verify the model, and the comparison with other traditional models furtherly proves the effectiveness and superiority of the model in this paper.

Automated summary

This paper proposes a surface roughness prediction model based on non-deformed chip thickness and generalized regression neural network in robotic belt grinding, which has significant application value.