Improving the accuration of train arrival detection based-on vibration signal using accelerometer sensor

Traffic accidents at railway crossings are increasing. One of the factors that causing accidents is human error. So, an automatic crossing gate system is needed. In previous work, a train arrival detection system has been built, but the system accuracy is only 65.72% and in the other research, the ADXL345 sensor is used because suitable to the application of railway engineering which has exact demands on sensitivity but only in the form of vibration monitoring. In this study, the train arrival detection system will be designed using an ADXL345 accelerometer sensor to increase the system accuracy from previous studies. The system will determine “condition 1” when the vertical axis value of the sensor reading is more than the maximum threshold value or less than the minimum threshold value. Condition 1 indicates that there is a train crossing and condition 0 is a condition when there is no train passing. Accuracy improvement is obtained by changing the threshold value in the previous study that uses the maximum and minimum values of the idle condition or one time peak-to-peak value of the three accelerometer sensor axes becomes the maximum value plus peak-to-peak from the idle condition and the minimum value minus peak-to-peak from the stationary condition or three times the peak-to-peak value on one axis with the vertical direction parallel to the earth’s gravity. Threshold determination is carried out based on three tests when the sensor is in an idle condition on the rail bearing to the three sensors used. Based on the values of -0.761670017 g for the maximum threshold and -0.949158001 g for the minimum threshold, the results of the designed system show the average accuracy in detecting passing train is 66.38%, the average accuracy when there is no passing train is 100%, so that the average accuracy of the system is 83.19% which means there is an increase in accuracy of 17.47% from previous studies.