Multidisciplinary Design Optimization from an Enterprise Perspective

Matteo Nicolich



Multi-disciplinary analysis has become complex due to the extensive use of different simulation tools and laboratory tests both at component and subsystem level. Issues mainly arise from the necessity to compare and share data having different formats, levels of fidelity and handling. Industries like automotive, today are challenged by numerous complex and often conflicting requirements such as compress vehicle design cycle time, lower the weight and cost of vehicles, and improve product performances, e.g., safety, NVH, durability, quality, fuel economy, reliability etc. To satisfy these stringent requirements, multidisciplinary design optimization (MDO) for large-scale vehicle program applications are run from early stages of design to fine tuning of single components.

Traditionally, MDO is conducted by a MDO subject matter expert who needs to collect all attribute information and models, integrate them into a large-scale optimization workflow, and solve the multi-objective optimization problem through an engineer-centric desktop solution. This is a time consuming, difficult, and engineer dependent process, which may not reach the full potential of the advanced MDO technologies. In this paper, we present a real customer case where SOMO, tightly integrated with private clouding computing systems, can make significant contributions for cross-attribute program developments in term of weight saving, fuel economy increase, meeting or exceeding multiple attribute requirements, as well as reducing product development time and engineering cost.

Large-scale vehicle designs MDO requires the handling of a great number of variables and becomes challenging using high fidelity models for direct optimization studies. Esteco’s enterprise technology, SOMO, enables users to create and manage a collaborative and distributed MDO simulation processes easily integrating and combining high-level fidelity simulations models and, at the same time, to create and manage highly sharable low fidelity models based on advanced meta-modeling techniques.
The combined use of multilevel models allows engineers to choose between going for a direct optimization analysis or quickly explore the feasible area using fast computing Response Surface Methodology (RSM) models and then verify few selected candidate solutions against the complex multidomain high fidelity models. The complete traceability of data, models and results is ensured throughout the entire simulation lifecycle for each vehicle program.