MODELLING OF DIFFERENT ASPECT OF THE CUTTING PROCESS BY USING ANNs

In addition to developing novel ways of part formation (e.g. for the purpose of rapid prototyping), cutting remained one the most important manufacturing technologies (2),(11). Though being the traditional way of manufacturing, the exact description of cutting processes is unknown, i.e. there are no comprehensive analytical models available for them (11). However, today's measuring and process technologies allow measuring the important process-related signals with high sampling rates. Consequently, it is possible to build up a database from the collected data and to produce appropriate process models by using novel technologies, such as artificial intelligence (AI) techniques. AI techniques have found their application in nearly every field (design, planning, control, optimisation, etc.) of manufacturing (6). Artificial neural networks (ANNs) can handle strong non-linearities, large number of parameters, missing information and the relationships between various parameters (5),(9). The cutting operation is very complex, therefore cutting is one of the typical fields of ANN based process modelling (5). ANN was used to find hidden relations in the set of cutting-related data as presented in (7), (8), (12). In the investigations to be described in the paper, the turning process was selected to analyse because of its relative simplicity. Different cutting situations were generated by adjusting machining parameters (variations of feed rate, depth of cut, cutting speed and cutting diameter) and measuring some physical parameters like force (3 dimension), surface irregularity, power, rubbing or built-up edge and temperature. Measured data were appropriately processed in order to generate a set of features for training the ANNs and some process related machining tolerances were determined.