The application of neural networks in fixture planning by pattern classification

Abstract

The objective of this paper is to study the fixture planning of modular fixtures for cutting by means of a computer-aided fixture system (CAFS). The fixture planning presented here integrates the database of modular fixtures, neural networks, the heuristic algorithm of fixture position and the use of Advanced Modelling Extension Release 2.1 (AME) of AutoCAD R12 as the 3D graphic interface. The function suitable for the AME environment was also controlled to obtain detailed data of the geometry and topology of the workpiece to develop the fixture planning for modular fixtures. First of all, the concept of group technology (GT) was used to establish the coding database of the modular fixture element for use in the system. This paper presents fixture planning which combines the pattern recognition capability of the neural networks and the concept of GT to group the workpieces with different patterns but identical fixture modes into the same group. After network training, the fixture mode of the workpiece to be clamped can be inferred, and the selection of the fixture elements can be completed.