Fast computation of inverse transient analysis for pipeline condition assessment via surrogate modeling with sparse sampling strategy

Inverse transient analysis (ITA) is a powerful tool for pipeline condition assessment, where the concerned parameters are decided by matching the measured transient signal with its physical model. This scheme is computationally expensive because, in the optimization procedure, the model has to be evaluated many times via time-domain full-wave numerical simulations. To address this problem, a fast algorithm of ITA is proposed, where the role of the numerical model is replaced by an easy-to-evaluate surrogate model. The Kriging method is employed to build the surrogate model of transient wave, given a series of samples of concerned parameters and the associated numerical model outputs. Considering the sparse nature of the high-dimensional pipe condition parameters, a sparse sampling strategy is proposed to optimize the samples of parameters. Two application examples, i.e., pipe viscoelastic parameter estimation and leakage identification, are introduced to demonstrate that the proposed method is as accurate as the traditional ITA but has a much lower computational cost.