Roller path solver system for multi-objective task-priority control of multipass conventional spinning

Abstract

A general design method for roller paths for multipass conventional spinning has not been established yet. The roller paths in a production site are often decided by experienced spinners by trial and error. In this paper, we propose a roller path solver system, which can automatically generate a roller path for spinning a cylindrical cup and simultaneously realize multiple target values of the spun product and forming process. On the basis of an artificial neural network trained with parameters of the roller paths and property values of the forming results, the path parameters that satisfy the target output are solved by an iterative algorithm using a pseudoinverse Jacobian matrix. In addition, the null space of the Jacobian matrix is used to handle the main and subtasks according to their priority. The experiments show that the target cup height, wall thickness, forming time, and surplus path length can be simultaneously achieved as the main task while moderately controlling the number of passes and the roller feed ratio as the subtask.

Keywords

• metal spinning
• incremental forming
• metal forming
• toolpath generation
• machine learning
• artificial neural network