Controlling Tolerence Stacks for Efficient Manufacturing

Datum-hierarchy (rooted) trees provide a highly efficient and sound basis for tolerance analysis and automated process planning. This paper builds on the previous work by the authors and shows that it is possible to optimise the structure of datum-hierarchy trees automatically. A process planner specifies an initial tree, which is improved upon by using an evolutionary algorithm and is checked against manufacturing and design constraints. A cost function that takes into account machine changes, datum changes and the sum of tolerance stacks drives the mutation process to generate "minimum cost" trees. Experiments were conducted on four industrial parts to demonstrate the feasibility of the approach. The results demonstrate that the method can reduce machine and datum changes and the sum of tolerance stacks. They also show that the order of priority of the cost criteria has a significant effect on the final shape of the trees.