The improvement of thermal error modeling and compensation on machine tools by CMAC neural network

In this paper, a cerebellar model articulation controller (CMAC) neural network is proposed for thermal error modeling in machine tools. The CMAC is a systematic learning algorithm which can search for the nonlinear and interaction characteristics between the thermal errors and temperature field on the machine tools. The CMAC is investigated in terms of accuracy in prediction, robustness to sensor placement, speed of learning, and tolerance to sensor failures. Experimental measurements of the spindle drift errors for both a horizontal machining center and a CNC turning center were performed using capacitance sensors and thermal sensors. Results show that the CMAC model has better performance than other modeling methods in robustness to sensor placement and speed of learning. This makes determination of the sensor locations easier, and reduces calibration time. In addition, a sensor failure detection algorithm is developed to provide better reliability.