基于APSO-Robust-ELM的大坝变形原始监测数据粗差识别方法

A Method for Identifying Gross Error in Raw Dam Deformation Monitoring Data Based on APSO-Robust-ELM

针对大坝原始监测数据普遍存在粗差的问题,提出了稳健估计与极限学习机相结合的粗差识别方法;并在此基础上,利用自适应粒子群算法寻找神经网络的最佳隐含层节点数;最后利用大坝安全预测模型验证所用方法的必要性和适用性。在工程实例分析中,将APSO-Robust-ELM法与Robust-ELM法、罗曼诺夫斯基准则和拉依达准则的处理结果进行对比,结果表明,APSO-Robust-ELM法能够更好的识别原始监测数据中的粗差,从而提高大坝安全运行的管理效率。

Aiming at the universal of gross error in the raw monitoring data of dams,a gross error identification method combining robust estimation and extreme learning machine is proposed,and on this basis,an adaptive particle swarm algorithm is utilized to find the optimal number of nodes in the hidden layer of the neural network.Finally,a dam safety prediction model is utilized to validate the necessity and applicability of the proposed method.In a study case,the processing results of the APSO-Robust-ELM method are compared with those of the Robust-ELM method,Romanovsky criterion and Pauta criterion,and the results show that the APSO-Robust-ELM method is able to better identify gross error in the raw monitoring data,thus improving the management efficiency of the safe operation of dams.