Production efficiency in metal forming processes can be improved by implementing robust optimization. In a robust optimization method, the material and process scatter are taken into account to predict and to minimize the product variability around the target mean. For this purpose, the scatter of input parameters are propagated to predict the product variability. Consequently, a design setting is selected at which product variation due to input scatter is minimized. If the minimum product variation is still higher than the specific tolerance, then the input noise must be adjusted accordingly. For example this means that materials with a tighter specification must be ordered, which often results in additional costs. In this article, an inverse robust optimization approach is presented to tailor the variation of material and process noise parameters based on the specified product tolerance. Both robust optimization and tailoring of material and process scatter are performed on the metamodel of an automotive part. Although the robust optimization method facilitates finding a design setting at which the product to product variation is minimized, the tighter product tolerance is only achievable by requiring less scatter of noise parameters. It is shown that the presented inverse approach is able to predict the required adjustment for each noise parameter to obtain the specified product tolerance. Additionally, the developed method can equally be used to relax material specifications and thus obtain the same product tolerance, ultimately resulting in a cheaper process. A strategy for updating the metamodel on a wider (noise) base is presented and implemented to obtain a larger noise scatter while maintaining the same product tolerance.

通过实施可靠的优化，可以提高金属成型工艺的生产效率。在稳健的优化方法中，考虑了材料和过程的分散，以预测和最小化目标均值附近的产品变异性。为此，传播输入参数的散布以预测乘积的可变性。因此，选择了一种设计设置，在该设置下，由于输入散射而导致的产品变化将降至最低。如果最小产品偏差仍高于特定公差，则必须相应地调整输入噪声。例如，这意味着必须订购规格更严格的材料，这通常会导致额外的成本。在这篇文章中，提出了一种鲁棒的逆向优化方法，可以根据指定的产品公差来调整材料和过程噪声参数的变化。在汽车零件的元模型上进行材料的稳健优化和定制以及工艺分散。尽管稳健的优化方法有助于找到最小化产品差异的设计设置，但只有通过要求较少的噪声参数分散，才能实现更严格的产品公差。结果表明，所提出的逆向方法能够预测每个噪声参数所需的调整，以获得指定的产品公差。另外，所开发的方法同样可以用于放松材料规格，从而获得相同的产品公差，最终导致更便宜的工艺。