Abstract:

Based on the multi phase flow simulation model and radial basis function artificial neural network surrogate model, sensitivity of design variables, which affected the surfactant enhanced remediation efficiency of NAPLs contaminated aquifer, was analyzed using the Sobol global sensitivity analysis method. The coefficients of determination were 0.977 8 and 0.981 6 when the number of training data were 12 and 22, which demonstrated that with the number of training data increasing, the approximation accuracy increased. The results of sensitivity analysis indicated that total pumping rates had the greatest percentage contribution to the remediation efficiency (total sensitivity was 0.491 2), and the next was remediation duration (total sensitivity was 0.468 5), surfactant concentration had the smallest  percentage contribution to the remediation efficiency (total sensitivity was 0.124 2). The interaction of variables also had contribution to the remediation efficiency, but the interaction effect was small. Sensitivity analysis results lay a foundation for optimization design of aquifer remediation.

Key words: sensitivity analysis, Sobol method, surrogate model, groundwater pollution, remediation