Recognition of the manufacturing features with computer integrated method of artificial neural networks

In this study, A system, recognizing 10 standard manufacturing features (slot, profile protrusion, pocket, cylindrical hole, boss, step, corner, blind hole, profile hole, blind slot), was developed by using Artificial Neural Networks (ANN) method. For this purpose, firstly, an interface was developed by Delphi 7.0 programming language. STEP file structure of product created with CAD software was automatically taken by using this interface. Through this interface, Input vectors required for Artificial Neural Networks were produced by analyzing STEP file structure. The input vectors were transferred to EasyNN-plus artificial neural networks package software. Artificial Neural Network was trained by using the input vectors and optimal artificial neural network model to able to maintain identification of manufacturing features was developed. Information of optimal artificial neural network model were presented to previously developed interface. Thus, the program was updated. The Input vectors, automatically produced by updated program, were processed and which predefined manufacturing features on the product was found to be. Finally, to test accuracy of the developed model, STEP files of product reserved were used. Manufacturing features on the product, allocated to test, were predicted correctly with a value over %80. As a result, this study constituted a step in order to ensure automation of design and manufacturing. Key Words : Step data structure, artificial neural networks, feature recognition.