Predictive Analysis of Railway Wheel Defects Using Artificial Intelligence

Abstract

: Railway wheel condition monitoring plays a critical role in ensuring operational safety and minimizing maintenance costs. Conventional inspection methods are o昀琀en time-consuming, labor-intensive, and ine昀케cient for early fault detection, highlighting the need for intelligent diagnostic solutions. 吀栀is study presents an Arti昀椀cial Intelligence (AI)-based predictive model that integrates fuzzy logic and Arti昀椀cial Neural Networks (ANN) to assess railway wheel conditions using vibration level, frequency, and train velocity as input parameters. 吀栀e ANN model was trained and evaluated using a dataset of 1,000 fuzzy logic-generated samples representing di昀昀erent railway wheel operating conditions. Wheel conditions were classi昀椀ed into four categories: Good (G), Low Damaged (LD), Faulty (F), and Dangerous (D). Among the evaluated training algorithms, the Levenberg–Marquardt (LM) algorithm achieved the highest prediction accuracy (97.5%). 吀栀e proposed framework demonstrates the potential of AI-assisted predictive maintenance for rapid and reliable railway wheel fault diagnosis. Although the current investigation is based on a simulated fuzzy-rule-based dataset, the developed model provides a foundation for future validation using real-world railway monitoring data to enhance practical deployment