EI、Scopus 收录
中文核心期刊
梁乃生, 脱友才, 邓云, 贾云霄. 基于主成分分析与支持向量机的渠道闸前冰输移与堆积判别模型[J]. 力学学报, 2021, 53(3): 703-713. DOI: 10.6052/0459-1879-20-391
引用本文: 梁乃生, 脱友才, 邓云, 贾云霄. 基于主成分分析与支持向量机的渠道闸前冰输移与堆积判别模型[J]. 力学学报, 2021, 53(3): 703-713. DOI: 10.6052/0459-1879-20-391
Liang Naisheng, Tuo Youcai, Deng Yun, Jia Yunxiao. CLASSIFICATION MODEL OF ICE TRANSPORT AND ACCUMULATION IN FRONT OF CHANNEL FLAT SLUICE BASED ON PCA-SVM[J]. Chinese Journal of Theoretical and Applied Mechanics, 2021, 53(3): 703-713. DOI: 10.6052/0459-1879-20-391
Citation: Liang Naisheng, Tuo Youcai, Deng Yun, Jia Yunxiao. CLASSIFICATION MODEL OF ICE TRANSPORT AND ACCUMULATION IN FRONT OF CHANNEL FLAT SLUICE BASED ON PCA-SVM[J]. Chinese Journal of Theoretical and Applied Mechanics, 2021, 53(3): 703-713. DOI: 10.6052/0459-1879-20-391

基于主成分分析与支持向量机的渠道闸前冰输移与堆积判别模型

CLASSIFICATION MODEL OF ICE TRANSPORT AND ACCUMULATION IN FRONT OF CHANNEL FLAT SLUICE BASED ON PCA-SVM

  • 摘要: 冬季渠道输水过程中浮冰容易在闸前形成堆积体,导致过水断面束窄,严重时影响渠道的输水效率和安全运行. 为判断渠道闸前浮冰的输移状态,开展了明渠水槽平板闸孔自由出流条件下的室内物理模型试验,提出了一种基于主成分分析与支持向量机(principal component analysis and support vectormachine,PCA-SVM)的闸前冰堆积与输移判别模型.通过相关性分析法确定输入特征间存在信息重叠,进而采用主成分分析法对特征向量进行降维,提取贡献率为86%的第一主成分和贡献率为7%的第二主成分作为输入特征,利用网格搜索方法确定多项式、高斯径向基和Sigmoid核函数的最优参数,通过混淆矩阵确定最优核函数为高斯径向基,最优核函数参数C为137,\gamma为0.37,建立PCA-SVM模型对试验数据进行监督学习.结果显示,模型在验证集上预测精确率为0.94,准确率为0.97,F1-Score为0.97,上游水流弗汝徳数(Fr_1)和闸前水流弗汝徳数(Fr_2)是渠道闸前冰输移与堆积的主要影响因素,闸孔相对开度(H/e)和闸门相对淹没水深(H_1/H)是次要影响因素.进一步将已建立的模型应用在倒虹吸口浮冰状态判别试验中,验证开发模型的分类性能.研究成果可为冬季输水渠道的调度管理和安全运行提供重要参考.

     

    Abstract: Floating ice is easy to formed an ice jam in front of flat sluice of channel water delivery during the ice-affected seasons, which will affect the efficiency and safe operation of the channel in severe cases. Here, an experimental for free outflow of flat sluice in the open channel were studied based on indoor physical model test. In order to judge the condition of ice float in front of gates of the water conveyance channel. A discriminant model of ice accumulation and entrainment in front of gate based on Principal Components Analysis-Support Vector Machine(PCA-SVM) algorithm is proposed. The correlation analysis method is used to determine the information correlation between input features, and then the PCA method is used to reduce the dimension of the feature vectors. The aim is to improve the computational performance of the model. The first principal component with a contribution rate of 86% and the second principal component with a contribution rate of 7% were extracted as input features. The optimal parameters of Polynomial kernel function (POL), Gaussian Radial basis kernel function (RBF) and Sigmoid kernel function (SIG) were determined by grid search method. The optimal kernel function was determined as RBF by confusion matrix, and the optimal kernel function parameters C were 137 and \gamma were 0.37. The PCA-SVM model was used for supervised learning of the experimental data. It is found that the Fr_1 and Fr_2 are the main influencing factors of ice entrainment or jam in front of gates, and the H/e and H_1/H are the secondary influencing factors. Furthermore, the established model was applied to the identify the floating ice state in front of the inverted siphon. The aim is to verify the classification performance of the developed model. The results are of the great value for the dispatching management and safe operation of water delivery channels during ice period.

     

/

返回文章
返回