Autonomously triggered model updates for self-learning thermal error compensation

Abstract

The presented method significantly increases the self-optimization ability of thermal error compensation models by triggering on-machine measurements when unknown thermal conditions occur. These conditions, which are not represented by the training data of the compensation models, are identified by a novelty detection approach based on one-class support vector machines. The results show that the autonomously triggered on-machine measurements applied to a 5-axis machine tool overcome the trade-off between precision and productivity for thermal error compensation. The non-productive time to detect an exceedance of the predefined tolerances is reduced by 78% without significantly reducing the precision of the thermal error compensation.