APPLICATION OF A NEURAL-NETWORK MACHINE LEARNING-MODEL IN THE SELECTION SYSTEM OF SHEET-METAL BENDING TOOLING

Sheet metal bending was one of the earliest manufacturing techniques t o be developed in the manufacturing industry. The machine specificatio ns such as pressure capacity, bending length and so on must be conside red in the selection of sheet metal bending tooling in order to reach the optimal choice based on the needs of the decision maker. This pape r proposes a model using machine learning from neural networks in an e xpert system of sheet metal bending tooling. In this system, the patte rn classification capability of the back-propagation neural networks i s utilized to complete, by means of training and testing of the networ ks, the knowledge acquisition and knowledge inference in the expert sy stem. In this learning model, the learning methods are divided into tw o categories according to their problem attributes: digital attributes and conditional attributes. The problem with two digital analytical a ttributes and the problem with multi-decision conditional attributes b oth yield ideal learning effects. In this paper, a learning model with conditional attributes is used to develop the expert system of sheet metal bending tooling selection, which belongs to a multiple decision model.