To objectively and accurately study the relationship between the grain size distribution characteristic of solid in deb-ris flow and the influential factors of debris flow, 23 debris flow gullies in Wudongde Reservoir region are further investigated and analyzed. 16 debris flow influential factors that have closely related to the debris flow's fluid characteristics are chosen, and the kernel principal component analysis ( KPCA) is used to reduce dimension of these influencing factors, and a linearly independent principal component is formed. The results show that largest prediction error is 2. 6%, which demonstrates that the influential factors have complicated mapping relation with each other, also possess well nonlinear correlation.%为客观准确研究泥石流固体堆积物粒度分布特征与泥石流影响因素之间的关系，通过对某水库库区的23条泥石流沟进行深入的调查研究分析，选取与泥石流流体性质有密切关联的16个泥石流影响因素，利用核主成分分析法( KPCA)对影响因素进行降维，形成线性无关的主成分。然后利用神经网络，预测了泥石流固体堆积物的分形维数。结果表明，网络预测误差最大仅为2.6%，由此说明泥石流各影响因素间存在着复杂映射关系，具有良好的非线性相关性，能决定固体堆积物的粒度分布特征。
展开▼
