基于未确知均值聚类分析的砂土地震液化评价

发布时间：2018-07-31 09:50

【摘要】：基于未确知测度理论,建立了砂土地震液化判别和液化势分级的未确知均值聚类分析模型和方法。针对砂土地震液化评价中的许多不确定性影响因素,选用地震震级,地面地震加速度幅值,标准贯入击数,比贯入阻力,砂土相对密实度,砂土平均粒径和场地地下水位等7个评价指标作为判别因子;选用17个砂土样本作为训练样本,建立各评价指标的未确知测度函数,以样本中的各评价指标数据的平均值表示其分类中心;利用相似权赋权方法确定评价指标的权重,依据未确知测度距离判别地震液化等级;根据建立的模型对训练样本回判,回判正确率为94.12%。将建立的模型对20个测试样本进行判别,将判别结果与地震液化的实际情况、BP神经网络和SOFM神经网络等方法的评价结果进行了对比。研究表明,该模型的评价结果与实测结果,以及BP神经网络、SOFM神经网络等方法的评判结果一致性较高。
[Abstract]:Based on the theory of unascertained measure, an unascertained mean clustering analysis model and method of sand soil liquefaction discrimination and liquefaction potential classification is established. In view of many uncertain factors in sandy soil seismic liquefaction evaluation, earthquake magnitude, ground earthquake acceleration amplitude, standard penetration number, ratio penetration resistance, sand relative density, sand 7 evaluation indexes, such as soil average particle size and ground water level, are used as discriminant factors, and 17 sand samples are selected as training samples to establish unascertained measure function of each evaluation index. The classification center is expressed by the average value of each evaluation index data in the sample, and the weight of evaluation index is determined by the method of weighting the similarity weight, according to the unascertained method. The knowledge measure distance is used to judge the grade of seismic liquefaction. According to the established model, the training sample is returned and the correct rate is 94.12%. to distinguish 20 test samples. The results are compared with the actual conditions of earthquake liquefaction, the BP neural network and the SOFM neural network. The evaluation results of the model are in good agreement with the measured results, as well as the results of BP neural network and SOFM neural network.
【作者单位】：中南林业科技大学土木工程与力学学院;
【基金】：湖南省重点学科建设项目(2013ZDXK006) 中南林业科技大学引进高层次人才科研启动基金(104-0094)
【分类号】：TU435

【相似文献】