BSCE 2020, magna cum laude, Undergraduate Research Project Title: Application of Artificial Neural Network to City-Scale Finite Element Model Generation

Abstract:

The main objective of the study is to develop an automated frame model generating tool that utilizes artificial neural networks (ANN). To train and test multiple topologies at once, the feed forward ANN algorithm was programmed, and enhanced by parallel computing and high performance computing (HPC). Different sets of ANN topologies were trained and tested using heuristic-based framing models to predict the column locations, and the floor elevations given the GIS data on building footprint and height, respectively. For each application, the optimal ANN topologies were identified, and integrated into the developed frame model generating tool. Using the developed tool, the three-dimensional (3D) frame models of target structures in Pateros, and San Juan were successfully created. Given the generated line element meshed models of the structures in the study areas, the city seismic response was estimated using an existing finite element analysis (FEA) module. The generated 3D frame models, and corresponding estimated seismic response of selected structures were verified. The application of ANN to city-scale finite element (FE) model explored in this study was properly documented for future research use.

Keywords: feed forward artificial neural network, city-scale finite element model generation, geographic information system, heuristic-based framing models, city seismic response

*This research project was co-advised with Mr. Karlo Daniel Q. Colegio  of Structural Engineering Group (SEG)