The student Xuban Telleria Ariztimuño obtained the qualification 'CUM LAUDE'.

The student Xuban Telleria Ariztimuño obtained the qualification 'CUM LAUDE'.

---

Thesis title: Development of a Novel Surrogate Modelling Technique and a Robust CAD-CAE Template Association Procedure for the Design and Assessment of Scalable Mechanical Product Families.

Court:

• Chairmanship: Germán Castillo López (Universidad de Málaga)
• Vocal:Jorge Dorribo Camba (Purdue University)
• Vocal:Malek Ben Salem (ANSYS)
• Vocal: Ibai Ulacia Garmendia (Mondragon Unibertsitatea)
• Secretary:Iñigo Llavori Osa (Mondragon Unibertsitatea)

Abstract:

Offering customised products is a strategy used by many companies to build competitive advantage. The development of such products is costly and time-consuming however, and manufacturers are constantly seeking new ways to streamline processes and reduce production costs. In this sense, the use of product families is an extended solution to efficiently develop different product references by modifying the product platform. Scale- based strategies are particularly widespread in the development of mechanical product families, since dimensions play an important role in the structural integrity and functionality of these kinds of products. However, in large families, many of the possible product references might never be required by the customer, rendering the time and money invested in the design process wasted. To prevent the unnecessary design of products, several companies follow the Modify To Order (MTO) design-delivery strategy, which consists of designing only customer requested products. However, in this design-delivery strategy all the design tasks performed after the Order Penetration Point (OPP) directly affect product delivery time. Regarding these design tasks, the two main factors that slow down the design process of mechanical product families are: (i) the iterative nature of the design process, and (ii) the verification time required to ensure the functionality and reliability of the ordered product.

The use of non-representative scaling rules in the design process of mechanical product families increases the number of design iterations. Although these scaling rules can be determined following the fundamentals of mechanics of materials, they are hard to correctly implement in complex geometries. Hence, surrogate modelling methods are frequently employed as an alternative solution to generate representative scaling rules for complex geometries. There exists a broad range of surrogate model types in the literature and each model can be configured in multiple ways. The choice of the surrogate model and its configuration parameters greatly affects the accuracy of the final result. In addition, the optimal model selection is highly conditioned by the case study in which it is employed and its non-linearity level. For this reason, several authors in the literature propose evaluating different types and configurations of surrogate models as a procedure to choose the most appropriate for each case. This process, however, is cumbersome and time consuming, particularly in the context of a fast-paced manufacturing environment.

As regards the second factor, CAE verification is frequently requested by the customer to ensure functionality and reliability requested mechanical product. For that purpose, it is common to use CAD-CAE templates to streamline the process, or going one step further, to define design automation applications. The geometrical entities required by the CAE-features to define the application zone of loads, boundary conditions, meshing rules, and contact zones, among others, are assigned to attributes in the CAD template to define a CAD-to-CAE information transference bridge. However, topology changes in the CAD template caused by the product platform scaling process usually corrupt the geometrical entity assignation of the attributes, generating inadequate CAE models. For this reason, additional CAD-CAE definition tasks are required, slowing down the design-analysis verification process and limiting the potential for implementing handoff design-analysis automated applications.

In this doctoral thesis, therefore, two techniques to streamline the development process of scalable mechanical product family references are proposed:

• A novel surrogate modelling method capable of adapting to any non-linearity level case with a single configuration which reduces the number of scaling deign iterations to near zero.
• A CAD template attribute definition procedure and a code-based solution to solve attribute corruption problems in geometry scaling processes, suppressing unnecessary manual CAD-CAE fixing tasks.

The proposed surrogate modelling technique, as well as the manual and code-based procedures were validated in different case studies:

• The proposed surrogate modelling technique was validated with 6 different case studies of different non-linearity levels (2 low, 2 medium, and 2 high). In addition, the proposed model was compared with other 14 surrogate model configurations of 8 well-known surrogate model types in the literature. The obtained results show that the error remained stable independent of the non-linear level of the case study, defining representative scaling rules.
• The manual and code-based procedures to prevent attribute corruption problems due to topology changes when scaling were validated in 6 case studies classified in three groups: (i) topology change due to dimensional modification, (ii) topology change due to CAD-feature pattern modification, and (iii) topology change due to the modification of the activation state of CAD-features. The results showed that following both proposed techniques, attributes were not corrupted when topology was altered.

Finally, the proposed techniques were implemented in the development process of a real valve family (SLAB API 17D family). The implementation of the techniques was validated as follows: (i) the mechanical behaviour of four valve references generated with the proposed surrogate modelling technique was evaluated with a high-fidelity FEM model, and (ii) the attribute corruption was analysed in five problematic cases identified in four valve reference cases. The obtained results of both validation procedures show that: (i) the determined scaling rules with the proposed surrogate model were able to efficiently design these valves reducing to cero the design iterations in the scaling process, and (ii) all the attribute corruption problems identified were solved with the proposed procedures.