Railway Track Monitoring System

Abstract

Listen

Translate article

For the protection and proximity of railway networks it is substantial to Promptly detect and identify faults in the railway tracks. In this paper, railway track fault diagnosis is approximated from the vertical and lateral acceleration using a MPU6050. MPU6050 consisting of three sensors namely gyroscope, magnetometer and accelerometer are used to distinguish line and level as symetricities in a railway track. A GSM module is used to notify the location of faults on tracks. Arduino Microcontroller is interfaced using Arduino UNO IDE. The results show that the condition of railway track irregularity and railway track striation can be approximated constructively. The processed data is uploaded to the open source cloud provider thingspeak.com. The use of various Machine Learning Algorithms are proposed to accomplish the above tasks based on the commonly available measured signals. By considering the signals from multiple railway tracks in a geographic location, faults are diagnosed from their spatial and temporal dependencies. The irregularities in the railway tracks are detected using the Inertial Monitoring Unit, providing the necessary data about future deformities using Machine Learning. Using Python 3.0, a generative model is developed to show that the AdaBoost network can learn these dependencies directly from the data. Seven different classification algorithms used for this project are Logistic regression,Naive Bayes Algorithm,Support Vector Machine, Ensemble Machine (Average) learning Algorithm, XGBoost Classifier, Extreme Machine Learning and AdaBoost Classifier. Among the above 7 classification algorithms, AdaBoost Learning has given the highest accuracy,i.e of 93.93 %. The AdaBoost Machine Learning Model is used throughout the model.