Abstract: The principle of sandiness water ablating pump impeller and its serious affects are analyzed.Based on complete overhaul log files data of Jiangnan Water Factory 1～6 pump sets,500VYM pumps produced by Ebara co.ltd.,of Japan, fuzzy support vector machine is adopted to forecast ablation performance index under limited sample data sets and complex factors.It can support pump sets duty time and overhaul time setting,human and outlay arrangement,spare parts purchase, and so on.Corresponding algorithm is derived and analyzed,and its experiment result is compared with that of the ANN (artificial neural network) and common kernel functions.The proposed algorithm has better performance and it can provide reference for processing this kind of problem.