In powder bed fusion (PBF) additive manufacturing, the definition of part orientation is one of the most important steps as it affects the quality, the cost and the build time of products. Different works already attempted to propose methodologies for the assessment of optimal build orientation based on criteria such as the minimization of support volume. Elicitation works with industry experts have shown that they use much more varied rules to determine the orientation of parts. For instance, they do not treat the different surfaces of the part the same way (e.g., experts state that “priority surfaces of the part must be oriented close to vertical”). Today, the available tools do not allow integrating these kind of specifications. This paper discusses a knowledge-based methodology for the evaluation of part candidate orientations in PBF. Desirability function approach is used to translate companies’ expertise in the form action rules into mathematical functions that are tested on geometries to provide metrics for assisting the decision-making. A case study is presented to illustrate the use of this desirability function approach on complex part orientation problem.

In reality, the behavior of processes is sometimes vague and the observed data is irregular. This research proposes an approach for optimizing fuzzy multiple responses using fuzzy regression. In this approach, each response repetition is transformed into a signal to noise ratio then modeled using statistical multiple regression. A trapezoidal fuzzy regression model is formulated for each response utilizing the statistical regression coefficients. The most desirable response values and the deviation function are determined for each response. Finally, four optimization models are formulated for the trapezoidal membership fuzzy number to obtain the optimal factor level at each number. Two case studies are adopted for illustration, where excluding response fuzziness will result in misleading optimal factor settings if solved by the traditional optimization techniques. In conclusion, the proposed approach based on fuzzy regression approach can successfully optimize fuzzy multiple responses in a wide range of manufacturing applications on the Taguchi's method. Moreover, compared to other approaches, such as data envelopment analysis and grey relational analysis, the proposed approach has the distinct advantage of being able to generate models using only a small number of experimental data sets and minimizing inherent variations.