Abstract:

In this paper, we continue a previous work where a framework for automated self-paced learning in CAD courses for Engineering Education was presented. In this particular work, we focus on automated grading of student CAD designs, which is a powerful tool very rarely encountered in Universities, in contrast with other areas where the approach has become commonplace. The challenge of automated evaluation in CAD, as in other STEM fields, is twofold: first, we must procedurally generate a set of exercises for homework that are substantially different from one student to the other, still maintaining the same degree of difficulty. The second problem to tackle is to automatically grade student submissions, taking into account that exercises are unequal, and that they might have been built in CAD environments via different sequences. While the procedural generation of exercises has been presented in a previous paper, here we focus on the last challenge. In order to evaluate automatically, an easily implementable novel numerical approach is presented. However, this approach should not be applied blindly, in order to avoid false positives and negatives in grading. The analysis of this aspect leads to the conclusion that procedural wording generation and automated grading are deeply intertwined and should be co-designed. The new approach has been applied to a large group of students in consecutive years, and an analysis of their behavior and performance is also presented.

Keywords:

Engineering Education, Automatic Evaluation, Internet Web-based Learning, Problem-based learning, Computer Aided Design, CAD.