Interval Prediction of Building Foundation Settlement Using Kernel Extreme Learning Machine

Abstract

Dynamic building foundation settlement subsidence threatens urban businesses and residential communities. In the temporal domain, building foundation settlement is often dynamic and requires real-time monitoring. Accurate quantification of the uncertainty of foundation settlement in the near future is essential to advanced risk management for buildings. Traditional models for predicting foundation settlement mostly utilize the point estimates approach, which provides a single value that can be close or distant from the actual one. However, such an estimation fails to quantify estimation uncertainties. The interval prediction, as an alternative, can provide a prediction interval for the ground settlement with high confidence bands. This study, proposes a lower upper bound estimation approach integrated with a kernel extreme learning machine to predict ground settlement levels with prediction intervals in the temporal domain. A revised objective function is proposed to further improve the interval prediction performance. In this study, the proposed method is compared to the artificial neural network and classical extreme learning machine. Building settlement data collected from Fuxing City, Liaoning Province in China was used to validate the proposed approach. The comparative results show that the proposed approach can construct superior prediction intervals for foundation settlement.