DECISION MECHANISM FOR WORKPIECE LOCATING DATUM BASED ON NEURAL NETWORK IN SETUP PLANNING

In this paper, the decision mechanism based on Neural Network, in which a new kind of numerical representation is used to aptly represent symbolic relationships among feature information about workpiece locating and setup planning, is used to to determine the workpiece locating datum. The idea is based on potential abilities of self learning and distributed data processing within ANN (Artificial Neural Network). In this ANN decision mechanism, 15 elements involving the virtual surrounding area and real area of face, ratio of length to width of the workpiece, area difference between two set ups, tool axial direction, dimension tolerance, geometric tolerance, surface finish, etc. are extracted as the input vector, while 3 elements corresponding to the primary, secondary and tertiary datum faces of 3 2 1 locating mode are used as the output vector. A sample prismatic workpiece which consists of 35 faces is examined by using this mechanism, and the test results show that this approach is able to generate locating knowledge that fulfils the intelligent interference process and fixture planning requirements that are commonly practiced by experienced planners, and is able to generate practicable solution of locating datum. It also can improve competence of nonlinear computing and fault tolerance for complex workpiece.