Adaptive group-wise modeling of thermally induced errors of a turning center

Abstract

Traditional multivariate regression analysis-based thermal error models use only one polynomial of several temperature variables to predict thermal errors, which will produce lower local prediction accuracy for a longer machining process with sudden changes of machining parameters, and hence the group-wise modeling method is proposed in this paper. Resorting to hard break points and soft break points, the grouping work is completed in two steps: hard grouping and soft grouping. The positions of hard break points are optimized using the genetic algorithm toolbox in Matlab software to realize adaptive grouping. The mechanism for updating the thermal error model coefficients vectors for different soft groups is developed. The modeling test is carried out on a turning center for which the positions of thermal key points are optimized. The prediction results for radial and axial thermal errors show that four hard break points can basically meet the requirements at the di value of 80%, so the group-wise modeling method is helpful to advance the prediction accuracy of thermal errors.