Rapid Assessment of Seismic Risk for Railway Bridges Based on Machine Learning

• DOI: 10.1142/s0219455424500561
• Publication Date: 2023-06-10T05:56:15Z
• Authors: Yong Huang; Jing He; Zhihui Zhu
• Keywords: Seismic risk; Bridge (graph theory); Vulnerability

When an earthquake occurs, railway bridges will suffer from different degrees of seismic damage, and it is necessary to assess the risk bridges. Unfortunately, majority studies were done on highway without taking into account bridge characteristics; hence they are not applicable Furthermore, current research methods for assessment cannot be performed quickly, problems subjective personal experience, complicated calculations, time-consuming. This paper we use machine learning damage prediction empirical vulnerability curves represent results, creating a rapid procedure. We gathered tallied data 335 that damaged in Tangshan Menyuan earthquakes, found six variables had substantial impact outcomes, categorized levels five categories. It essentially multi-classification problem. In order solve this problem, four algorithms tested: Random Forest (RF) Back Propagation Artiifcial Neural Network (BP-ANN), PSO-Support Vector Machine (PSO-SVM), K Nearest Neighbor (KNN). was RF most effective method, with accuracy rate up 93.31% training set 89.39% test set. Then study describes new procedure detail rapidly assessing 269 chosen at random sample pool. Firstly, collated, then rating predicted using RF, finally curve drawn two-parameter normal distribution function purpose assessment. The study’s findings can used as guide choosing approach its inputs build model