Data Science based Multidisciplinary Design Optimization (MDO) methods are disrupting the way the Computer Aided Engineering (CAE) analyses is performed. Traditional Response Surface based optimization methods fail to handle huge design variables and responses that are involved in the real-world vehicle design optimization problems. The challenge in deployment of MDO algorithms in solving real-world automotive structural design problems is huge time associated in running full vehicle Finite Element Models to solve highly non-linear complex phenomenon involving geometric, material and contact nonlinearities with large number of design variables and multiple performance responses and constraints pertaining to multiple domains namely, durability, crash and NVH. With the availability of powerful workstations and the advanced CAE tools, it has become possible to generate huge set of simulation data sets, considering practically any number of design variables that exist in a real-world vehicle design problem. Initially, huge data sets are obtained by performing simulations on a particular vehicle architecture, predictive models were then developed using the data sets. Considering gages of the vehicle structures as design variables and durability requirements, front-end intrusions during an IIHS offset impact test, the modal frequencies of the critical structural members as the constraint variables and using the developed predictive models, optimization is performed for Lightweighting the structural members. It is noted that, using the current methodology, a significant reduction in the simulation time and considerable weight saving of nearly 20% is obtained.
