Enhanced interactive parallel coordinates using machine learning and uncertainty propagation for engineering design

Abstract

The design process of an engineering system requires thorough consideration of varied specifications, each with potentially large number of dimensions. The sheer volume of data, as well as its complexity, can overwhelm the designer and obscure vital information. Visualisation of big data can mitigate the issue of information overload but static display can suffer from overplotting. To tackle the issue of overplotting and cluttered data, we present an interactive and touch-screen capable visualisation toolkit that combines Parallel Coordinates and Scatter Plot approaches for managing multidimensional engineering design data. As engineering projects require a multitude of varied software to handle the various aspects of the design process, the combined datasets often do not have an underlying mathematical model. We address this issue by enhancing our visualisation software with Machine Learning methods which also facilitate further insights into the data. Furthermore, various software within the engineering design cycle produce information of different level of fidelity (accuracy and trustworthiness), as well as with different speed. The induced uncertainty is also considered and modelled in the synthetic dataset and is also presented in an interactive way. This paper describes a new visualisation software package and demonstrates its functionality on a complex aircraft systems design dataset.