A Data-Based Model of the Thermo-Elastic TCP Error Using the Encoder Difference and Neural Networks

Abstract

AbstractThe thermo-elastic tool center point (TCP) error has been an ongoing research focus, due to its large effect on the workpiece quality. Existing models to compute the thermo-elastic TCP error already perform quite well regarding the accuracy and speed of computation. However, the models are often time consuming in their parameterization, expensive to apply or are error-prone due to the used model inputs. The work presented in this paper addresses these issues by introducing the encoder difference as model input. Since the encoder difference easy and inexpensive to measure, it yields a high potential for industrial use. Therefore, in this paper, the correlation between the encoder difference and the thermo-elastic TCP error is investigated. Since the physical relationship between the encoder difference and the thermo-elastic TCP error is complex, it is necessary to use an artificial neural network to compute the resulting TCP error. Due to the variety of artificial neural network (ANN) types, with different capabilities, a range of different networks is tested regarding their capability to compute the thermo-elastic TCP error. To conclude the paper, a method to parametrize such models is derived from the gathered results.