Session: 02-09-01: Session #1: Variation Simulation and Design for Assembly Description

Paper Number: 88775

88775 - Coupling Sampling-Based Tolerance-Cost Optimization and Selective Assembly – An Integrated Approach for Optimal Tolerance Allocation 

Optimization has gained increased attention in product development and is nowadays profitably used to solve complex design problems. In tolerance design, tolerance-cost optimization makes use of its systematic and efficient approach to allocate part tolerance values in a cost-optimal way. As the method is an important link between design and manufacturing, various aspects from manufacturing have gradually been frontloaded into the design-driven methods of tolerance allocation to represent given manufacturing conditions as realistically as possible at an early stage. The usage of sampling-based tolerance analysis techniques thereby enables the consideration of machine-specific part tolerance distributions as well as individual batch sizes. In doing so, tolerance-cost optimization is not restricted to the selection of one single machine to realize an allocated tolerance, but it rather supports a concurrent allocation of tolerance values and distribution of the total batch on multiple machines. With the aim to further exploit this potential, this article presents a novel approach combining tolerance-cost optimization and optimal selective assembly. Since the integration of multiple machine selection already divides the total batch into several individual bins, this information can be used to optimally match these and selectively assemble the corresponding parts. An initial discussion and exemplary study illustrate that this will lead to less non-conform and scrapped assemblies, which indirectly offers the potential to widen the individual part tolerances and thus to reduce the total tolerance-related manufacturing costs. To address this aspect adequately in tolerance-cost optimization, an efficient solution is needed to additionally find the optimal mating bins, characterized by the individually chosen batch sizes and tolerance values. Hence, a global optimization problem is introduced to simultaneously identify the best combination of tolerance values, machines with its batch sizes and sorting orders of the individual batches. The latter requires, first, the identification of all possible combinations and, second, their mathematical description in a way that they can be suitably treated by the optimizer as a set of independent design variables. This consequently leads to a complex, mixed-integer optimization problem, as the number of possible machine combinations is finite and discrete. The optimization problem is solved with the aid of metaheuristic optimization algorithms, which have already proven their suitability for sampling-based tolerance cost optimization to handle the complexity of the highly non-linear and stochastic problems. An exemplary application of the developed method to an academic use case confirms the theoretical findings and indicates that it can reveal hidden cost potentials. A final discussion on the potentials and remaining challenges intends to identify future research directions fostering the ongoing shift to a manufacturing and assembly process-oriented sampling-based tolerance-cost optimization.

Presenting Author: Martin Roth Friedrich-Alexander-Universität Erlangen-Nürnberg

Presenting Author Biography: Martin Roth is a research associate and PhD student in the research group ‘dimensional management’ at the Institute of Engineering Design of the Friedrich-Alexander-Universität Erlangen-Nürnberg. In his current research, he focuses on virtual product development and computer-aided tolerancing. His area of expertise is the usage of optimisation techniques for least-cost tolerance design, which is also known under the term tolerance-cost optimisation.

Authors:

Martin Roth Friedrich-Alexander-Universität Erlangen-Nürnberg
Markus Johannes Seitz Friedrich-Alexander-Universität Erlangen-Nürnberg
Benjamin Schleich Friedrich-Alexander-Universität Erlangen-Nürnberg
Sandro Wartzack Friedrich-Alexander-Universität Erlangen-Nürnberg