AUTHOR=Zhang Youdong , Xiao Guijian , Gao Hui , Zhu Bao , Huang Yun , Li Wei 

TITLE=Roughness Prediction and Performance Analysis of Data-Driven Superalloy Belt Grinding

JOURNAL=Frontiers in Materials

VOLUME=Volume 9 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/materials/articles/10.3389/fmats.2022.765401

DOI=10.3389/fmats.2022.765401

ISSN=2296-8016

ABSTRACT=In grinding the hard processing material such as high temperature alloy, have improve burn lower surface defect and surface integrity has a unique advantage, however, due to the randomness and uncontrollable distribution of abrasive particles, grinding heat is high, the processing characteristics of flutter, the nonlinear relationship between processing parameters and the surface quality becomes a crux of roughness prediction. In this paper, a prediction method for grinding roughness of GH4169 superalloy belt is proposed based on neural network algorithm. The belt grinding system of superalloy is introduced first. According to the experience formula of process parameters and surface roughness, a neural network prediction model of process parameters and surface roughness was established. Secondly, bionic genetic algorithm and particle swarm optimization algorithm were used to replace weights and thresholds in BP neural network algorithm with randomly generated weights and thresholds, so as to obtain the global optimal solution and avoid the local optimal solution in algorithm iterative solution. Finally, the average absolute error (MAE) of BP algorithm, GA-BP algorithm and PSO-BP algorithm is 0.12, 0.085 and 0.078 respectively. The goodness of fit (R) of BP, GA-BP and PSO-BP were 0.954, 0.988 and 0.993, respectively.