PSI - Issue 64

Eshwar Kumar Ramasetti et al. / Procedia Structural Integrity 64 (2024) 557–564 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

561

5

3. Artificial Intelligence and Neural Network Classifier 3.1. Classification Approach

The primary goal of this work is to develop generic AI models for classifying vehicle movement on the bridge with the constraint that the bridge is operated to have only one direction of vehicle movement. The sensor’s data of acceleration signals were evaluated in time intervals of 30 seconds to get the feature values which are used for the classification. The features chosen in this work are: (i) maximum acceleration; (ii) minimum acceleration; (iii) true RMS value; and (iv) standard deviation, as well as (v) the acceleration value. Based on these features, a binary classification to detect vehicles passing on the bridge was developed. Furthermore, the vehicles were classified as a car, a truck, or a large truck in a second step as a multi-label classification. The data was labeled by selecting thresholds for the maximum acceleration value based on the existing dataset. Therefore, the noise generated by the sensors while there was no vehicle movement on the bridge was measured. By this, the threshold value for the acceleration signal without any vehicle movement on the bridge was set to 0.02 m/s 2 ; if it was greater than 0.02 m/s 2 , a vehicle had passed, and if it was less than 0.02 m/s 2 , no vehicle movement occurred. The threshold value for cars is greater than 0.02 m/s 2 and less than 0.2 m/s 2 ; for trucks, it is greater than 0.2 m/s 2 and less than 0.4 m/s 2 ; and for large trucks, it is greater than 0.4 m/s 2 . The basic classification approach for vehicle movement is depicted in Fig. 5 below. Fig. 6 illustrates the labelling of the dataset for the feature maximum acceleration with different colors for the different classes. There is noise generated from the bridge and sensors due to the environmental conditions, and it was assumed to be 0.02 m/s 2 , which is marked with the blue color. The movement of cars is indicated by orange with an acceleration value above 0.02 m/s 2 and less than 0.2 m/s 2 , trucks in red with an acceleration value above 0.2 m/s 2 and less than 0.4 m/s 2 , and large trucks in black with an acceleration value above 0.4 m/s 2 .

Fig. 5. Classification approach for vehicle movement