Comparison of a Deep Learning-based Axle Load Estimator and the Matrix Method in Strain Gauge-based Bridge Weigh-In-Motion Systems

One aspect of structural health monitoring of bridges is to monitor vehicular traffic. The demand for moni-toring bridge performance and life cycle has led to measuring traffic flow. We have previously developed a deep learning-based axle load estimator showing promising results considering COST 323 benchmark. This paper compares the deep learning-based solution to the established matrix method under different circumstances. Results show that the deep learning-based solution achieves better accuracy on several datasets of the BME-Simulated I corpus and has a better ability to handle noise than the matrix method.