山地学报
山地學報
산지학보
JOURNAL OF MOUNTAIN SCIENCE
2015年
3期
374-378
,共5页
边坡变形%变形预测%最小二乘支持向量机%量子粒子群优化
邊坡變形%變形預測%最小二乘支持嚮量機%量子粒子群優化
변파변형%변형예측%최소이승지지향량궤%양자입자군우화
slope displacement%prediction of deformation%least square support vector machine%quantum-behaved particle swarm optimization
滑坡变形受外界影响因素作用的机理十分复杂,难以采用简单方法对其进行预测,因此建立快速准确的滑坡预测模型十分重要。采用比一般支持向量机( SVM)预测效果更好且计算速度更快的最小二乘支持向量机( LSSVM)方法,选用RBF核函数对边坡位移时序数据进行训练和预测,并引入量子粒子群算法( QPSO)对LSSVM模型参数γ和σ进行全局寻优,避免了人为选择参数的盲目性,提高了模型的预测精度。将优化模型应用于新滩滑坡和卧龙寺新滑坡的变形预测,并与传统的LSSVM、PSO - LSSVM模型进行预测精度及收敛性对比分析。结果表明,QPSO - LSSVM模型较传统方法在预测精度上有了明显提高,且收敛速度明显加快,说明QPSO - LSSVM模型在边坡位移时序预测中具有良好的应用价值。
滑坡變形受外界影響因素作用的機理十分複雜,難以採用簡單方法對其進行預測,因此建立快速準確的滑坡預測模型十分重要。採用比一般支持嚮量機( SVM)預測效果更好且計算速度更快的最小二乘支持嚮量機( LSSVM)方法,選用RBF覈函數對邊坡位移時序數據進行訓練和預測,併引入量子粒子群算法( QPSO)對LSSVM模型參數γ和σ進行全跼尋優,避免瞭人為選擇參數的盲目性,提高瞭模型的預測精度。將優化模型應用于新灘滑坡和臥龍寺新滑坡的變形預測,併與傳統的LSSVM、PSO - LSSVM模型進行預測精度及收斂性對比分析。結果錶明,QPSO - LSSVM模型較傳統方法在預測精度上有瞭明顯提高,且收斂速度明顯加快,說明QPSO - LSSVM模型在邊坡位移時序預測中具有良好的應用價值。
활파변형수외계영향인소작용적궤리십분복잡,난이채용간단방법대기진행예측,인차건립쾌속준학적활파예측모형십분중요。채용비일반지지향량궤( SVM)예측효과경호차계산속도경쾌적최소이승지지향량궤( LSSVM)방법,선용RBF핵함수대변파위이시서수거진행훈련화예측,병인입양자입자군산법( QPSO)대LSSVM모형삼수γ화σ진행전국심우,피면료인위선택삼수적맹목성,제고료모형적예측정도。장우화모형응용우신탄활파화와룡사신활파적변형예측,병여전통적LSSVM、PSO - LSSVM모형진행예측정도급수렴성대비분석。결과표명,QPSO - LSSVM모형교전통방법재예측정도상유료명현제고,차수렴속도명현가쾌,설명QPSO - LSSVM모형재변파위이시서예측중구유량호적응용개치。
The mechanism of slope deformation is complicated,because it is influenced by outside factors. It is dif-ficult to adopt simple method to predict,so establish a fast and accurate slope displacement prediction model is very important. The method of least squares support vector machines( LSSVM)with higher accuracy than standard sup-port vector machines method is used to train and simulation the slope displacement-time data. And the quantum-be-haved particle swarm optimization( QPSO)is adopted to optimize the parameters(γ,σ)of LSSVM model in order to avoid artificial arbitrariness and enhance the forecast accuracy. For comparison,the model of QPSO-LSSVM, LSSVM and the traditional SVM are used to forecast the same series displacement-time data of Xintan slope and Wolongsi slope. The results indicate that the QPSO -LSSVM method is much better than traditional method in terms of forecast accuracy and can be well applied to the forecast of displacement-time series.