Abstract: To improve the accuracy of risk assessment for levee engineering, we present a multi-risk index evaluation approach for levee engineering based on extreme learning machine. First of all, 28 evaluation indices that affect the levee risk are comprehensively considered, and the analytic hierarchy process is adopted to establish a multi-risk index evaluation system which comprises of early warning system, levee engineering system, environmental system and social economic system. Subsequently, the extreme learning machine algorithm is employed for standardized processing of the 28 indices and constructing grading standards. With the risk indices and the grading membership as the input and output, respectively, the risk levels are divided, the evaluation indices are quantified, the values of multiple risk evaluation indices are estimated, and the risk severity is judged. With Kangshan levee, a key levee of Poyang Lake, as a case study, the present multi-risk evaluation index system is established and the values of multi-risk evaluation indices are computed using the extreme learning machine algorithm. The evaluation results demonstrate that the Kangshan levee is safe in general, which is in line with the actual situation of the Kangshan levee after two reinforcements. The reliability and effectiveness of the proposed method is verified by comparison with other methods. The proposed approach is expected to be applied to the risk assessment of other important hydraulic structures.

Key words: levee engineering, risk assessment, analytic hierarchy process, extreme learning machine, multi-index system