Integration of Machine Learning and Visualization for Effective Design Exploration

The world of CAE is experiencing continued growth in the number of variables involved in a Design of Experiments, either from combining various domains or expanding the complexity of models.

Engineers are now facing the task of analysing not only the sensitivity and robustness of the solution with respect to hundreds of individual components and parameters (both simulated and external), but must also be able to gain a holistic understanding of their interplay and effects on the results. With ever-tightening design cycles engineers are hard-pressed to quickly extract actionable information from simulation studies and balance dozens of competing requirements, constraints and optimization goals.

In this paper we explore how Machine Learning methods can be brought in to respond to this challenge by using visual predictive analytics for analysing and post-processing simulation results.

A platform that turns Machine Learning algorithms (such as feature selection, classification and regression) into interactive visuals can establish a basis for guided, interactive exploration of the design space.

We demonstrate the applicability of Machine Learning and Visualization methods in the field of industrial product design and development by using concrete examples from a variety of disciplines.