Intelligent prediction model based on genetic algorithm and support vector machine for evaluation of mining-induced building damage

Abstract

Characteristics of factors influencing mining-induced building damage are diverse, nonlinear, and multi-linear. For a better description of these factors, an intelligent prediction model for building damage induced by underground mining is developed based on the support vector machine (SVM). Based on a comprehensive consideration of geological, mining, and building factors, 10 factors are carefully selected. In particular, the mining-induced damage grade of the brick-concrete building structure is used as the main input variable in the proposed model. The damage grade and largest crack width of the brickconcrete building structure are selected as output variables in the proposed model. A total of 32 typical cases of mining-induced building damage in China are collected and used as training data. The radial basis function (RBF) is used for SVM classification and the application of the largest-crack-width regression model. To improve the model’s generalizability and predictive capacity, the genetic algorithm (GA) is adopted to select effective parameters for the SVM model, and then the corresponding identification of six group samples is performed. The classification and regression results show that the proposed prediction model using GA-SVM can predict the mining-induced damage of a brick-concrete building structure, and the evaluation results show good agreement with monitored data. This suggests the practicality of the proposed model in a wide range of engineering problems.