Sensor Fusion of Laser and Inertial Units with Kalman-KMeans-Fuzzy Framework for Real-Time Railway Geometry Monitoring

Maintaining railway track geometry integrity is essential to ensuring transportation safety and predictive maintenance. Conventional manual inspection methods are limited by low sampling frequency, subjective interpretation, and delayed anomaly detection. This study introduces a real-time, embedded monitoring system using VL53L0X infrared laser sensors and an MPU6050 IMU to measure gauge, cross-level height, and inclination. Sensors are mounted on a lightweight aluminum trolley and sampled every 0.5 seconds using an Arduino-based platform. A Kalman Filter reduces measurement noise, with tuned covariance matrices based on field calibration. Filtered outputs are clustered via K-Means (K = 2), validated by the Elbow Method and Silhouette Score (>0.6). Maintenance categories are assigned through a fuzzy logic system, with a ±1 mm sensitivity analysis confirming >85% decision stability. Field results demonstrate a measurement noise, achieving RMSE and MAE values of 0.8165 mm and 0.3175 mm for gauge and height, and 0.3086° and 0.0952° for inclination, respectively and a SNR gain from 0.5 dB to 21.7 dB. The low-cost, modular setup supports scalable, condition-based maintenance and demonstrates robustness in noisy environments. This approach offers a practical foundation for future integration with predictive analytics and digital twin technologies in smart rail infrastructure.