Predictive data analytic approaches for characterizing design behaviors in design-build-fly aerospace and aeronautical capstone design courses

Abstract

Predictive data models and interactive visualizations can be highly effective in understanding workload and skills assignment issues within design-build-fly teams in the aerospace industry. Capturing data that is needed to build predictive models in usable forms and then subsequently applying appropriate data mining techniques to derive insights from such data is a significant challenge. The ultimate goal of our work is to understand design behaviors among engineers that can lead to cost reductions and expediting product development in extremely complex engineering environments. The present study, pioneered by a large US aerospace company working with educators at 5 major engineering schools in the US, engineering education researchers, and practicing engineers, is a first step towards achieving this overall vision. In this paper, we characterize how engineering students enrolled in a senior capstone course interact and perform on complex engineering tasks commonly seen in the aerospace industry. We describe our instrumentation methodology and the data architecture for an associated analytics platform. We use course clickstreams, social networking and collaborations as the basis for our observations.