Under the circumstances of increasing number of standard machine parts the parts selection becomes more important than ever before. Selection of appropriate bearings also plays an important role. The purpose of this paper is to establish an appropriate model for the selection of bearings, which is essential in the preliminary phase of machine design.
Typical decision-making approaches are compared in terms of selection characteristics: The expert system is proper to the problem that has small number of attributes, and/or requires logical and persistent knowledge. The multi-attribute decision-making is recommendable to the cases in which determination of attributes is easy, and/or that requires reliable and numerical knowledge. The artificial neural network can be successfully used when it is not easy to determine common attributes and to obtain explicit knowledge, or it is easy to collect credible cases.
A stepwise selection model has been suggested according to the above comparison results. In this model an artificial neural network trained with design cases is used to select a bearing mechanism as a first step. Then the selection of bearing sub-type is performed using the weighting sum model. Finally, types of peripherals such as lubrication methods are determined by a rule-based expert system. Experimental selections of bearings show that the model is compatible with the bearing design problem.
Finally, an integrated system is designed in order to develop a selection model as a system made of artificial intelligence techniques and multi-attribute decision-making. Stepwise bearing selection is performed by means of collaboration among the artificial neural network, weighting sum model, and rule-based expert system. In the system several commercial software packages has been loosely coupled.