考虑变形因子模式下基于正则化极限学习机的大坝变形预报方法

Dam Deformation Prediction Method Based on Regularized Extreme Learning Machine Model under Considering Deformation Factors Mode

本文将结构风险最小化原则引入极限学习机模型,建立了在考虑变形因子模式下大坝变形预报的正则化极限学习机模型。该模型不仅计算速度较快,而且具有较强的泛化能力。通过对实际工程监测数据的详细分析,结果表明正则化极限学习机模型可以避免原极限学习机模型会导致过学习现象发生的可能,且其预报精度要优于原极限学习机模型、支持向量机模型与BP神经网络模型。显示了将其应用于大坝变形数据分析与预报领域是完全行之有效的。

The structure risk minimization principle was introduced to extreme learning machine （ELM） model. Based on this, regularized extreme learning machine model of dam deformation prediction under considering deformation factors mode was built. Not only calculation speed of the model is very fast, but also generalization ability is very strong. By analyzing the results of an engineering example in detail, the results show that regularized ELM model can avoid the possibility of original ELM model which over learning phenomenon happened. Moreover, prediction precision of regularized ELM model is su and BP neural network model. So, regularized ELM model analysis and prediction.