Selection of Input and Output Variables for Ann Based Modeling of Cutting Processes

Modeling of manufacturing operations is an important tool for production planning, optimization and control. Artificial neural networks (ANNs) can handle strong non-linearity, large number of parameters, missing information. Based on their inherent learning capabilities ANNs can adapt themselves to changes of the production environment and can be used also in the case if there is no exact knowledge about the relationships between various parameters of manufacturing. Typical field of ANN based operation modeling is cutting. The relationships of the physical phenomena incorporated into the cutting operation are very complex. In the application of these models several tasks can be determined: e.g. ? in the planning phase the surface roughness is predefined and the model is expected to select the cutting parameters and to predict the cutting force, while ? during supervised production the cutting parameters are known and e.g. the cutting force is measured and the produced surface roughness is to be estimated. In the above assignments, the operation parameters are the same but the operation model has other variables on the input and on the output sides. In this paper a method is presented to build a general operation model with the requested accuracy. This method incorporates: ? determination of the number of output variables ? determination for every parameter to be input or output The method is also useful in the case of strong nonlinear relationships. Experiments show the applicability of the approach.