CONCEPT MODELING TECHNIQUES FOR THE DESIGN OF AUTOMOTIVE STRUCTURES

During the last decades the overall trend in the automotive industry has been to shorten the design cycles and decrease the production costs while increasing the product quality with respect to competitors. In such a context there is an ever growing need for analysis that leads the design process from the concept phase onward. On the other hand, a conventional Finite Element (FE) model of a vehicle can be created only when its detailed Computer-Aided Design (CAD) model is available, which automatically excludes obtaining early stage simulation results. In this sense novel Computer-Aided Engineering (CAE) methodologies are required to support the concept modeling and optimization of vehicles. This challenge defines the main framework of the research in the present dissertation. The leading motivation is to introduce new improvements and developments in the field of CAE concept approaches. The focus is put on the design of the car structure. The main objective is to obtain in an efficient way accurate early stage predictions of its static and low frequency dynamic behavior by means of FE concept models. Predecessor-based concept modeling methods, which start from the reference FE model of an existing car and aim at achieving variant or incremental improvements of it, are addressed. Developments in regard to two main groups of techniques are introduced: methods based on simplified FE models of the vehicle structure and methods based on mesh morphing. The related achievements are presented in this dissertation. To begin with, the current challenges of 1D beam concept modeling have been thoroughly investigated. Guidelines have been given on good practices to overcome the intrinsic limitations of the existing techniques and to make them more accurate and reliable. Beam Bounding Box has been proposed as a novel approach for 1D beam concept modeling and optimization handling, which is accurate, computationally beneficial, easy to implement and to apply. Furthermore superelement joints have been introduced as means for the creation of more accurate simplified FE models. A major breakthrough has been achieved in the field of sizing optimization by identifying, improving, implementing and validating successfully Differential Evolution (DE) as an advanced alternative to the state-of-the-art gradient-based methods. Finally, surrogate modeling based on mesh morphing of predecessor FE models has been introduced as another option to enable fast modification and optimization studies in the concept stage. The added value of all these contributions in the automotive engineering practice has been demonstrated by their application on a number of realistic industrial case-studies throughout the dissertation.