A Machine-Learning-Based Software for the Simulation of Regional Characteristic Ground Motion

Ground-motion simulations provide input time history data required for designing and assessing structures; however, the conducted by currently available tools only match design spectrum without verifying if statistical characteristics of duration are satisfied. A ground-motion simulation software was developed to resolve these issues. The employs machine learning methods amplitude, spectrum, features target region. Principal component analysis is employed extract from actual database detect characteristic ground motions predict acceleration response duration, based on prediction equations. results show that simulated motion can well. Therefore, more reasonable structure. Moreover, provides visualization functions enable user determine area obtain amplitude field intuitively.