Neural network models have been used in the recent past for performance prediction in machining operations.While the predictions are encouraging with comparable quantitative accuracy with conventional models,there is a need to predict performance 'on line'.If performance features such as various force components,vibrations and tool wear in cylindrical turning need to be estimated,on-line,there is no conentional model available to cater for this need.the conventional mechanics of cutting models are efficient to the extent of predicting individual performance but cannot estimate on line performance features.Empirical approach also cannot cater for 'on line' estimation of performance.In turning,tool-workpiece interference causes the wear growth and the failure.A turning tool can have network model is proposed for cutting tool wear detection,on-line,in a production environment.Extensive turning experiments on lathe is carried out covering a comprehensive range of cutting conditions.Using neural network architecture intelligent sensorintegration approach for identification of wear growth is developed.It is shown that the forces and vibrations as inputs to the neural network architecture hav esuccessfully predicted the progressive tool wear.
展开▼
