Bài báo khoa học

This study aims to predict the remaining design resistance and bending stiffness of steel column base plates considering metal corrosion, which helps construction managers make proper de- cisions about maintenance, reinforcement, or demolition. To achieve this, the paper introduces a method for predicting the design resistance and bending stiffness of the steel column base plate, considering metal corrosion using a hybrid Genetic Algorithm-Artificial Neural Network (GA- ANN) model. The GA-ANN model was trained using a dataset of 808 observations derived from calculating design resistance and bending stiffness of the steel column base plates. Fifteen input parameters of the GA-ANN model included geometric parameters and material properties of the concrete foundation block, base plate, steel column, and anchor bolts. Two output parameters were the design resistance and bending stiffness. The predicted results were then compared with those of the traditional Levenberg-Maquard-Artificial Neural Network (LM-ANN) model. Statis- tical indices including R2, RMSE, and a20 −index demonstrated that the GA-ANN model was superior in predicting capability. By integrating the training results of GA-ANN with a metal corrosion model, this study proposed a procedure for forecasting the remaining design resistance and bending stiffness of corroded steel column base plates. Additionally, the study evaluated the remaining design resistance and bending stiffness of corroded steel column base plates after 100 years. A free-access graphical user interface was developed to aid in the practical prediction process.