2023 ASEE Annual Conference & Exposition

The ever-increasing utilization of machine learning (ML) in technical fields suggests that educators should consider incorporating ML content into engineering and engineering technology curricula. This paper explores a course designed to instruct students on project-based machine learning in predictive maintenance. The course centered on a NASA dataset created for predictive maintenance exercises: a collection of simulated turbofans that were run until failure. A class of nine students was instructed to predict the remaining useful life of these turbofan units using various analysis techniques. Data processing and regression models were created in Google Colab via Tensorflow, Sklearn, and Pandas modules. The course began with classical regression approaches such as linear regression and then progressed to ML methods including neural networks, Long Short Term Memory networks, and random forests. Data processing and feature generation were also covered, as well as model design considerations such as hyperparameter searches. Student performance was evaluated with a self-efficacy survey conducted on the first and last day of the course. Participants began with low self-efficacy in knowledge and skill domains, but high attitudes regarding ML. By the end of the course, knowledge and skills saw a significant increase in score, with attitudes remaining constant. Students noted that they quickly understood the concepts and theory surrounding ML but struggled with coding and implementation. This course provides insight into the gains in ML knowledge and skills for non-CS students. The course also provides a pedagogical example that engineering and engineering technology instructors can employ to incorporate ML content into their courses. Data is presented to show that engineering students can develop practical ML skills for engineering applications.