MAPPING OF PROBE PRETRAVEL IN DIMENSIONAL MEASUREMENTS USING NEURAL NETWORKS COMPUTATIONAL TECHNIQUE

Pretravel is a major error source in dimensional measurements on coord inate measuring machine (CMM) and machine tools using trigger touch pr obes in modern manufacturing systems. The direction-dependent characte ristic of pretravel needs effective and efficient computational soluti ons. In this paper, a back-propagation neural network is designed and implemented to establish a pretravel map for touch trigger probes. Exp erimental pretravel information from a CMM is used to train the networ k, and the network is used to predict pretravel in various probe appro ach directions. Performance of the trained neural network has been eva luated and effective pretravel predication is demonstrated. Thanks to the parallel-computing nature of neural network technique, pretravel m apping can be implemented in real time once the off-line-trained netwo rk is obtained. The proposed approach can be incorporated into the mac hine systems to provide real-time pretravel correction in dimensional measurement processes. We also reported the advantages of neural netwo rk technology over the traditional method used in this application. (C ) 1997 Elsevier Science B.V.