Natural Language Processing Applications in Mechanical Engineering Education

Abstract NLP, or Natural Language Processing, is a branch of artificial intelligence, enabling machines to understand and respond to human language in both written and spoken forms. These systems are versatile, capable of tasks such as storytelling to resolving math problems and even producing code and images. Through continued research and development, the efficacy and efficiency of NLP solutions in addressing complex engineering challenges are anticipated to further improve. This will foster a new era of innovation at the intersection of artificial intelligence and engineering disciplines. Considering the practical utility of NLP in solving mechanical engineering problems, a pedagogical shift in educational paradigms to seamlessly incorporate NLP into mechanical engineering curricula seems inevitable. In this study we will analyze the current efficiency of NLP tools to solve mechanical engineering problems and propose a list of pedagogical approaches to incorporate the most efficient ones in mechanical engineering education. In this study, we evaluated various NLP tools such as ChatGPT-3.5, ChatGPT-4, Claude, Gemini, and Copilot. We used them to solve typical problems in mechanical engineering to assess their reliability and accuracy. A wide range of problems, such as vibration and Finite Element analysis were considered. We observed that many NLP systems struggled to provide detailed and precise answers in their default setups, presenting a significant challenge in their effective application for engineering problem-solving. To address this issue, we employed prompt engineering techniques to refine the responses generated by NLP systems. By utilizing methodologies such as defining user profiles and specifying desired outcomes, we aimed to enhance the relevance and precision of the solutions produced. While these adjustments led to gradual improvements in performance, we did not achieve considerable improvement. Next, we focused on developing and training a tailored NLP model specifically designed to tackle problems related to mechanical vibrations. By training this specialized model with relevant textbooks and study materials, we significantly boosted its proficiency in accurately handling complex vibration assignments and improved its performance for solid mechanics assignments. Furthermore, integration of NLPs with popular software platforms like MATLAB and Word through custom plugins to simplify the deployment of NLP solutions within engineering workflows were analyzed. Despite these plugins being in their early development stages, they encountered challenges in efficiently delivering solutions. In summary, this study successfully demonstrated the adoption and integration of NLP tools to solve engineering problems, particularly in mechanical engineering, demonstrating the potential for these technologies to revolutionize learning paradigms and problem-solving methodologies. At the conclusion of this study, we offered a list of various pedagogical approaches for integrating NLP effectively into mechanical engineering education. In the future we plan to implement these approaches in the classroom and study their effectiveness.