Mechanix: An Intelligent Web Interface for Automatic Grading of Sketched Free-Body Diagrams

Early engineering courses at large universities enroll hundreds of students, and as of now are taught in an in person, hybrid, or online setting. These large classroom size, as well as impersonal nature of the current safe teaching methods, makes it difficult for instructors to provide meaningful feedback on assignments that students complete. This lack of feedback has an effect on the development of early engineering skills that students develop, including the free-body diagrams (FBD). Even before the shift in current instruction methods, there was a growing concern amongst engineering educators that student's ability to idealize real world situations into these FBDs has been under-developed. Most existing homework methods do not provide students with impactful feedback on FBDs that may be draw during the solution of a problem, if any feedback is provided at all. Because of this concern, a Sketch-recognition based tutoring system called Mechanix has been developed by the Sketch Recognition Lab at Texas A&M specifically to provide real time tutoring in drawing Truss Systems. The application provides a drawing surface for students to hand-draw solutions as part of the submission process, as if they were submitting a problem through a pen-and-paper submission method. AI algorithms working in the background of the application identify the shape of the drawn FBD, the perceived internal definition of the shape, and any additional features added to the sketch. These algorithms then determine if the user inputs for the assignment are correct and will provide real time iterative feedback as to why a user's input may be incorrect. Results of past usage of Mechanix has shown positive results in increasing engagement in struggling students, while also be as effective as other traditional homework methods for teaching statics and dynamics concepts. This paper focuses on the current effect of Mechanix on instruction across the 5 universities involved in the study in pre and post Covid-19 instruction methods. The study uses a range of instruments to gauge student understanding through concept inventories, select homework assignments, exam grades, and specialized problem sets given to students to determine the impact of the application when compared to traditional homework methods already used by the participating schools.