Employing Machine Learning for Predicting Bridge Failure

Machine Learning is being used for several classification problems. Stochastic and statistical computing is being explored to derive conclusive decisions where human intervention is difficult in time and resource constrained situations. One such situation is bridge failures in cases of seismic impacts. In case of earthquakes, it is necessary to immediately evaluate the possibility of damage to bridges as they are critically important to carry out relief operations while carrying population and essential goods. However, human inspection in earthquake stricken areas may take a lot of time increasing the risk of using bridges which are severely damaged thereby risking human life. Hence, quick automated tools are required which can predict bridge damages quickly and based on less number of parameters with relatively high accuracy. This work presents a back propagation based neural network architecture for bridge failure prediction. The data set used is the Stanford Earthquake Dataset (STEAD). It has been shown that the proposed work attains high classification accuracy and low computation complexity making the model effective for quick evaluation of bridges from seismic impacts.