Development of an AI-Based Simple Harmonic Motion Classification System to Enhance Students’ Understanding of Vibration Concepts and Data Analysis Skills

Abstract

Understanding simple harmonic motion (SHM) is fundamental in physics, yet students often struggle with abstract concepts and connecting theory to experiments. This study developed an AI-based system to classify SHM patterns and evaluated its effectiveness in improving students’ conceptual understanding and data analysis skills. A Research and Development (R&D) approach adapted from Borg and Gall (1983) was used, comprising seven stages: information collection, planning, product development, expert validation, revision, small-scale trial, and classroom implementation. The system employed supervised machine learning to classify SHM into ideal, damped, overdamped, and critical damping. Data were collected using a modified physical pendulum equipped with a position sensor. Instruments included a conceptual understanding test, data analysis assessment, and student perception questionnaire, all validated for reliability. Descriptive statistics and normalized gain (N-gain) measured learning improvements. The AI system achieved 86% accuracy in classifying oscillation patterns. Pre-test and post-test results showed significant gains, with N-gain scores of 0.57 for SHM concepts, 0.56 for pattern identification, and 0.70 for data analysis. The integration of AI allowed rapid identification of different damping types, provided immediate feedback, and linked theoretical concepts with experimental observations. Combining AI-assisted classification with hands-on experiments enhanced students’ conceptual mastery, analytical skills, and engagement in physics learning. The study contributes theoretically by bridging abstract SHM concepts and observable phenomena. Practically, it provides an interactive, data-driven learning tool. Future research should expand the system to other physics topics, include larger and more diverse populations, and explore advanced AI techniques to further optimize learning outcomes and generalizability