Utilización de modelos de redes neuronales artificiales para predecir la influencia del tipo de fresado en la calidad del producto

Abstract

During the milling process of pieces with complex

surfaces, the choice of different machining

strategies suggested by the computer-aided

manufacturing software (CAM) leads to

deviations from the piece machined with respect

to the designed ideal surface. The knowledge

of the deviations generated with respect to the

final geometry of the piece allows to develop

correction modules in the software based on the

different machining strategies, enabling so the

executor to generate appropriate corrections

before manufacture, so that the finished products

will be within the design specifications. At the

same time, the application of artificial neural

networks (ANN) has been studied as a solution

to non-linear problems and other conflicting

parameters in machining process.

This paper evaluates the influence in the

geometry and surface finish of three different

milling strategies suggested by CAM software in

the manufacture of a product through ANN, thus

generating a set of helpful answers both from the

point of view of the analysis and prediction.