中国科技论文
中國科技論文
중국과기논문
Sciencepaper Online
2015年
15期
1772-1777
,共6页
尤军%殷耀国%李辉%柳永全
尤軍%慇耀國%李輝%柳永全
우군%은요국%리휘%류영전
沉降监测预报%MATLAB%曲线拟合%人工神经网络
沉降鑑測預報%MATLAB%麯線擬閤%人工神經網絡
침강감측예보%MATLAB%곡선의합%인공신경망락
settlement observation and prediction%MATLAB%curvfitting%artifical neural network (ANN)
沉降监测预报需要准确、及时和有效。针对沉降监测预报模型种类较多,导致数据处理和预报模型选用的随意性较大,提出了沉降监测预报模型的优选。以宁夏3个深基坑工程实际监测数据为案例,分别采用 MATLAB 中5种曲线拟合模型和3种人工神经网络模型进行预报,通过对比优选出适合的沉降监测预报模型。结果表明:GR(general regression)神经网络模型预报效果最好,BP(back propagation)神经网络模型和 RBF(radial basis function)神经网络模型预报效果较好;BP 神经网络模型的沉降预报精度比 RBF 神经网络模型稍高;三次样条插值模型可以进行沉降预报,但预报效果不及神经网络模型;采用不同模型进行联合预报,可以增强预报的可靠性;在工程实践中,发挥人工神经网络3种模型的预报优势,进行周期性预报和实时安全评价,具有一定实际意义。
沉降鑑測預報需要準確、及時和有效。針對沉降鑑測預報模型種類較多,導緻數據處理和預報模型選用的隨意性較大,提齣瞭沉降鑑測預報模型的優選。以寧夏3箇深基坑工程實際鑑測數據為案例,分彆採用 MATLAB 中5種麯線擬閤模型和3種人工神經網絡模型進行預報,通過對比優選齣適閤的沉降鑑測預報模型。結果錶明:GR(general regression)神經網絡模型預報效果最好,BP(back propagation)神經網絡模型和 RBF(radial basis function)神經網絡模型預報效果較好;BP 神經網絡模型的沉降預報精度比 RBF 神經網絡模型稍高;三次樣條插值模型可以進行沉降預報,但預報效果不及神經網絡模型;採用不同模型進行聯閤預報,可以增彊預報的可靠性;在工程實踐中,髮揮人工神經網絡3種模型的預報優勢,進行週期性預報和實時安全評價,具有一定實際意義。
침강감측예보수요준학、급시화유효。침대침강감측예보모형충류교다,도치수거처리화예보모형선용적수의성교대,제출료침강감측예보모형적우선。이저하3개심기갱공정실제감측수거위안례,분별채용 MATLAB 중5충곡선의합모형화3충인공신경망락모형진행예보,통과대비우선출괄합적침강감측예보모형。결과표명:GR(general regression)신경망락모형예보효과최호,BP(back propagation)신경망락모형화 RBF(radial basis function)신경망락모형예보효과교호;BP 신경망락모형적침강예보정도비 RBF 신경망락모형초고;삼차양조삽치모형가이진행침강예보,단예보효과불급신경망락모형;채용불동모형진행연합예보,가이증강예보적가고성;재공정실천중,발휘인공신경망락3충모형적예보우세,진행주기성예보화실시안전평개,구유일정실제의의。
Settlement observation and prediction requiress accuracy and timeliness in order to be effective.To amalgamate the many types of settlement observation and prediction modeling,which result in an arbitrarily large array of data processing and forecasting model options,a method for the optimization of the model of settlement monitoring and forecasting is proposed.Three case study examples of deep foundation pits in Ningxia,China were used in five kinds of MATLAB curve fitting models and three kinds of artificial neural network models in order to predict,by contrast,the optimized model base for settlement observation. The results show that the general regression (GR)neural network model predicted best,that the back propagation (BP)neural network and radial base function (RBF)neural network models forecasted better,that the BP neural network model’s accuracy for settlement observation was better than the RBF neural network model’s and that the cubic spline interpolation model can be settlement prediction but its forecast effect is not as good as neural network model’s.Also,the results indicate that using differ-ent models for joint forecasting can enhance the reliability of forecasts.In the practice of engineering,the capacity for prediction in the three artificial neural network models,as well as their periodic forecasting and real-time safety assessment,has practical significance.