CAJ | 학술논문

为了提高高速公路沉降预测的精度，提出了“双曲线＋神经网络”的混合模型对高速公路沉降进行预测．基于某省某高速公路K57＋580 M断面2003-2013年的实测沉降数据，采用双曲线模型对施工期和通车期分别进行沉降预测．然后利用构建的混合模型重新对该断面进行施工期和通车期的沉降预测，分析比较2种模型的预测精度．为了验证2种模型的预测精度，选取该高速公路K156＋100 M断面实测数据进行计算．结果表明，混合模型在施工期和通车期的预测值的中误差分别达到1．13和7．30 mm，预测精度相比双曲线模型在施工期和通车期分别提高了66．2％和41．8％．
위료제고고속공로침강예측적정도，제출료“쌍곡선＋신경망락”적혼합모형대고속공로침강진행예측．기우모성모고속공로K57＋580 M단면2003-2013년적실측침강수거，채용쌍곡선모형대시공기화통차기분별진행침강예측．연후이용구건적혼합모형중신대해단면진행시공기화통차기적침강예측，분석비교2충모형적예측정도．위료험증2충모형적예측정도，선취해고속공로K156＋100 M단면실측수거진행계산．결과표명，혼합모형재시공기화통차기적예측치적중오차분별체도1．13화7．30 mm，예측정도상비쌍곡선모형재시공기화통차기분별제고료66．2％화41．8％．
In order to improve the accuracy of subsidence prediction, a hybrid model based on neural network and hyperbolic is proposed for highway subsidence prediction.The used data is the actual observed settlement for the K57+580 M section in a highway project of one province from 2003 to 2013 .The data is calculated by using the hyperbolic model and the hybrid model in the construction period and the operation period, respectively.The prediction accuracies of the two models are ana-lyzed and compared.In order to verify the prediction accuracies of the models, the highway meas-ured data of another section;i.e., the K156+100 M section is selected for calculation.The results show that the prediction error values of the hybrid model in the construction period and the operation period can reach 1.13 and 7.30 mm, respectively and the accuracy is improved by 66.2% and 41.8%, respectively, compared with the hyperbolic model.