以向家坝水电站过机泥沙为研究对象，采用激光粒度仪分析过机泥沙粒径级配的体积百分比，运用多重分形理论对粒径分布的特征规律进行研究。结果表明：过机泥沙粒径具有良好的多重分形特性，结合多重分形理论，粒径分布整体平均上的不均匀大小为D1（B2）＞D1（C1）＞D1（C2）＞D1（A2）＞D1（B1）＞D1（A1）＞D1（D2）＞D1（D1）；[JP2]存在某一阈值q，使得当q≥5时，最大子集粒径分布的不均匀大小为D≥5(B2)＞D≥5(B1)＞D≥5(D2)＞D≥5(C2)＞D≥5(A2)＞D≥5(C1)＞D≥5(A1)＞D≥5(D1)，当q≤-7时，最小子集粒径分布的不均匀大小为D≤-7(C1)＞D≤-7(B2)＞D≤-7(A1)＞D≤-7(D1)＞D≤-7(C2)＞D≤-7(A2)＞D≤-7(D2)＞D≤-7(B1)（其中A1、B1、C1、D1为左岸机组过机泥沙样品编号，A2、B2、C2、D2为右岸机组过机泥沙样品编号）；多重分形奇异谱谱宽Δα应分为最小子集和最大子集奇异谱谱宽Δα1和Δα2分别计算；结合广义维数谱和多重分形奇异谱，关于粒径分布在局部上的不均匀程度，两者的结论是相同的；结合相关性分析，泥沙细化和最大子集分布不均匀大小呈现良好的负相关性，而泥沙细化和最小子集分布不均匀大小无关。

The power generation efficiency problem and hydropower stations turbine abrasion overwhelmed water conservancy engineers,especially during the flood season,when heavy rains,and the inflow and sediment concentration have reached their peaks.The power generation efficiency of hydropower stations is the best during this period,while the flow components of the turbine are also subject to abrasion.The Xiangjiaba Hydropower Station is located in the lower reaches of the Jinsha River.After its completion,it exerts tremendous flood control and power generation benefits.As the core equipment of the power station,the performance of the hydraulic turbine and the reliability of its operation are some of the keys to whether it can exert engineering benefits.To understand the rules of damage to hydraulic turbines by sediment in time,and to study various protective measures to reduce the abrasion damage to the surface of the turbine′s flowpassing parts by the sand-containing water flow,it is necessary to carry out the medium– and long–term monitoring of the sediment of the typical units on the left and right banks of the Xiangjiaba Hydropower Station to provide data support for the subsequent development of hydraulic turbine protection measures. The Xiangjiaba Hydropower Station flood season from June to September is taken as the research object.The maximum sediment concentration is selected twice a month by taking into account the time difference of the sediment particle size distribution during the flood season.The silt is sampled by the machine.For each sample,the left and right bank typical units are selected and enter the designated water intake inside the turbine volute for water sample collection.A total of 8 samples were taken during the entire flood season.After sampling the laser particle size analyzer was used to analyze the volume percentage of the machinesilt particle size gradation.The characteristic law of particle size distribution was studied using multifractal analysis. The results show that the particle size of the machined sediment has good multifractal characteristics.Combined with the multifractal theory,the overall average inhomogeneity of the particle size distribution is D1（B2）＞D1（C1）＞D1（C2）＞D1（A2）＞D1（B1）＞D1（A1）＞D1（D2）＞D1（D1）.There is a certain threshold,so that when,the uneven size of the largest subset particle size distribution is D≥5(B2)＞D≥5(B1)＞D≥5(D2)＞D≥5(C2)＞D≥5(A2)＞D≥5(C1)＞D≥5(A1)＞D≥(D1),at that time,the unevensize of the smallest subset particle size distribution isD≤-7，D≤-7(C1)＞D≤-7(B2)＞D≤-7(A1)＞D≤-7(D1)＞D≤-7(C2)＞D≤-7(A2)＞D≤-7(D2)＞D≤-7(B1)(A1,B1,C1,D1 are the sediment sample numbers of the left bank unit,and A2,B2,C2,D2 is the sediment sample number of the right bank unit).Combined with the generalized dimensional spectrum,the nonuniform distribution of the largest subset and the smallest subset of the sediment size of the Xiangjiaba Hydropower Station can not be effectively compared.Combined with the multifractal singularity,the spectrum width is Δα1 and Δα2,and the uneven size of the largest subset particle size distribution is Δα1 (B2)>Δα1 (B1)>Δα1(D2)>Δα1 (C2)>Δα1 (A2)>Δα1 (C1)>Δα1 (A1)>Δα1 (D1).The smallest subset particle size distribution is Δα2 (C1)>Δα2 (B2)>Δα2 (A1)>Δα2 (D1)>Δα2 (C2)>Δα2 (A2)>Δα2 (D2)>Δα2 (B1). Combining the generalized dimensional spectrum and the multifractal singular spectrum,the two conclusions are the same regarding the local unevenness of the particle size distribution.Combined with the correlation analysis,the sediment refinement and the maximum subset distribution unevenness show good results.The negative correlation of the silt refinement has nothing to do with the uneven size of the smallest subset distribution.