Abstract: Software based on neural network (NN) to compensate machine tools thermal deformation is studied. The knowledge acquisition which plays very important part in modeling a neural network is presented. Two kinds of machine tools are analyzed and the methods of learning data acquisition from experiment, especially temperature measurement and the choice of the number of temperature sensors are discussed. The learning data organization technique, including inductive and deductive learning is introduced. The results of neural network prediction and accuracy evaluation are given.

Key words: Error compensation, Knowledge acquisition, Neural network, Thermal deformation