Computer Applications in Engineering Education最新文献

Recent studies suggest that working with ChatGPT can generate more creative outcomes than humans alone. However, does ChatGPT retain its creative edge when humans have access to alternative information sources, such as Google Search or a human peer. This study addressed this question through a quasi-experiment with 230 Chinese university students in three groups (the human–ChatGPT group, the human–Google group, and human–human dyads) and a mathematical creativity task. The results showed that the human–ChatGPT group generated the most flexible and fluent solutions, while the human–human group produced the most original solutions in solving mathematical creativity tasks. The human–human group accurately assessed their actual performance, while the human–ChatGPT group overestimated it, and the human–Google group underestimated it. Furthermore, the study revealed that the human–Google group encountered greater difficulties, invested more effort, and reported lower levels of interest compared to the human–human and human–ChatGPT groups. Students found Google less useful than ChatGPT and their human peers. Similarly, students also found Google less effective than ChatGPT and their peers in enhancing self-efficacy. These findings highlight the benefits of human–human and human–ChatGPT co-creation in fostering mathematical creativity and call for further research on how to combine human inspiration with ChatGPT support to enhance it.

Program Outcomes (POs) are critical for engineering program accreditation, yet traditional evaluation methods often lack objectivity, consistency, and timely feedback. While machine learning (ML) has been applied to predict general student success, its use for predicting PO attainment levels from early academic data remains underexplored. This study introduces an integrated framework for computer engineering programs, combining a systematic PO assessment model with ML-driven prediction. The assessment model quantifies PO attainment rates (POAR) from weighted course assessments, mappings between Course Learning Outcomes (CLOs) and POs, CLO-assessment relationships, and student grades. Using these POARs, various ML techniques were trained on historical data from 327 graduates, utilizing their grades from 25 early-semester courses and graduation POARs. Our findings demonstrate that POARs can be successfully predicted from this early data, achieving a mean absolute percentage error around 5%. Consequently, this study presents a scalable and objective tool that (1) provides a systematic framework for POAR measurement; (2) offers an effective ML model for predicting graduation POARs of students; and (3) delivers data-driven insights for proactive student support, timely interventions, and evidence-based curriculum optimization, thereby supporting continuous program improvement and accreditation efforts.

When performing mathematical modeling, engineering education primarily focuses on understanding first principles to represent a phenomenon or process. With the advent of Machine Learning (ML), data-driven approaches to mathematical models have disrupted and challenged these traditional teaching/learning approaches. Data interpretability captures different dimensions, since engineers seek accurate predictions, causation, and analyze the interaction effects of process variables when modeling. While the effects of interaction effects have been previously taught using regression techniques, complex datasets might require employing alternative methods to precisely capture the complexity and nonlinear behavior. In this study, we present the conscious design of a novel teaching and learning approach for data-driven modeling, using a case study of the degradation of lithium-ion batteries to illustrate the interaction effects in modeling. We have selected there different interaction effects approaches when modeling: a regression model, exploratory data analysis, and ML. A validation and preassessment of the proposed teaching strategy were conducted to enhance the preparation and implementation of an in-class session, including strategies for its classroom integration. Our approach is innovative within the undergraduate engineering education context, since it introduces and highlights the significance of interaction effects to enhance students' abilities to interpret data, and think critically. This approach is totally reproducible, may be applied across other engineering disciplines, and has practical implications that could lead to its potential assimilation and utilization in industry.

Educational mobile game applications can effectively support learning by offering engaging ways to present difficult concepts. This is especially beneficial for teaching secondary school Geometry, which many students find challenging. This paper introduces Pythagorean Academy, a mobile app aligned with the Greek junior high school curriculum. The app not only tests students' understanding but also provides corrective feedback and theory explanations—features uncommon in educational apps. It also accommodates students with visual impairments and learning disabilities, such as Attention Deficit Hyperactivity Disorder, through tailored visuals and audio support. Incorporating gamification elements, the app boosts motivation and promotes autonomous learning, making Geometry more accessible and enjoyable. Statistical methods, including the Mann–Whitney U-test and the Wilcoxon signed rank test, evaluate the app's effectiveness. This study demonstrates how integrating Geometry concepts into a gaming framework can leverage modern technology to improve Geometry education.

3D CAD modeling technology has become an essential tool for product design across various industries, including machinery, aerospace, automotive, architecture, and healthcare. Consequently, numerous educational institutions offer training programs and certification exams to enhance and evaluate the modeling proficiency of 3D CAD system users. However, the manual grading process currently employed in 3D CAD modeling exams reveals several limitations, such as excessive time and effort, and challenges in maintaining consistency in evaluations. In mechanical CAD systems, in particular, users can create the same model using different features, making precise grading criteria essential. Additionally, the lack of self-directed learning capabilities among learners has emerged as a pressing issue, highlighting the need for more effective educational solutions. To address these challenges, this study introduces CADuBoost, an automated grading and feedback system for 3D CAD modeling education in mechanical engineering. CADuBoost compares student-submitted 3D CAD models with reference models through a comprehensive evaluation framework that processes both geometric and non-geometric data. Shape evaluation is conducted using neutral formats such as STEP and STL through point cloud comparison, multi-view image analysis, and dimensional accuracy measurement. Non-geometric evaluation is performed by extracting and analyzing design history and constraint information via the 3D CAD system's API. Furthermore, by providing visual feedback through color-coded geometric differences and detailed design history analysis, the system delivers personalized feedback that effectively fosters self-directed learning. The effectiveness of CADuBoost was validated through experiments in real educational settings, showing possibilities to improving students' modeling proficiency and self-directed learning abilities. This system is expected to enhance instructors' efficiency and improve the overall quality of education.

Human advancement hinges on the capacity to acquire knowledge and engage with complex ideas. Education, therefore, plays a pivotal role in shaping cognitive and societal growth. However, the increasing commercialization of education has raised significant concerns regarding declining academic standards, reduced student performance, and escalating unemployment. To address these systemic challenges, this study proposes a machine learning-based framework for predicting and evaluating Course Learning Outcomes (CLOs) and Program Learning Outcomes (PLOs) in an undergraduate engineering context. The proposed model analyzes historical academic records to investigate the influence of midterm and final assessments on overall grade performance and CLO/PLO attainment. Results indicate that CLO 1 has consistently achieved approximately 90% success over the past 2 academic years, a trend expected to persist based on predictive insights. These findings offer actionable guidance for academic departments to implement targeted interventions, such as scenario-based evaluations, to enhance student learning outcomes. By leveraging Python-based machine learning techniques, institutions can scale their data-driven assessment strategies and reinforce evidence-based educational practices. This study contributes to the growing field of AI-enhanced education, offering practical implications for improving academic quality and institutional decision-making.

Artificial intelligence (AI) is increasingly recognized as a vital tool in electrical engineering, offering automation, error reduction, and enhanced accessibility. However, its adoption has lagged compared to other fields, highlighting a need for a comprehensive examination of its applications and challenges. This study systematically reviews AI applications in electrical engineering, classifying research findings to uncover progress, challenges, and opportunities. It aims to identify trends, gaps, and implications to guide future research and practical applications. A systematic literature review (SLR) was conducted, analyzing studies published between 2014 and 2024. Fifty-seven publications meeting inclusion criteria were categorized into five themes: AI algorithms, power engineering, smart grid technologies, electric vehicle systems, and AI integration. The review revealed growing interest in AI applications within electrical engineering, with a significant rise in publications, particularly from China. AI algorithms demonstrated broad applicability and versatility across various domains, highlighting their potential for innovation. Additionally, there is a considerable opportunity for developing and applying frameworks to test AI innovations in electrical engineering. AI integration in electrical engineering has advanced significantly in areas such as power engineering, smart grid technologies, and electric vehicle systems. However, substantial untapped potential remains, particularly in developing frameworks for testing AI innovations. This review underscores the importance of global research efforts and identifies promising directions for advancing AI applications in electrical engineering research and practice.

The soil arching effect is a key concept in soil mechanics education. It is widely recognized as an important principle in geotechnical engineering, characterized by stress redistribution due to relative soil displacement, which impacts the safety and stability of geotechnical structures. Despite advances in classical theories and numerical methods, the complexity of models and formulas still presents significant challenges for students and engineers in understanding and application. To address this challenge, this study introduces a practical and educational solution by developing a computer-aided calculation platform for the soil arching effect, designed by Hunan Provincial Engineering Research Center of Advanced Technology and Intelligent Equipment for Underground Space Development in Hunan University, aimed at enhancing soil mechanics education through an intuitive MATLAB graphical user interface. The primary contribution of this study is the development of a platform that integrates seven theoretical models, enabling users to calculate key parameters, such as the soil arching ratio, by inputting soil properties and unloading width. The platform features real-time data visualization and interactivity, allowing users to easily select models, input parameters, and obtain results quickly, thereby facilitating comparative analysis across different theoretical frameworks. Compared to conventional teaching methods, the platform simplifies complex calculations and deepens students’ understanding of the soil arching effect. Results from student surveys indicate a remarkable improvement in comprehension and analytical skills, with high satisfaction regarding the platform's usability and educational value.