🤖 AI Summary
To address the insufficient accuracy of surrogate models in estimating probabilities of rare events—such as system failure—under uncertainty propagation, this paper proposes a heuristic sampling strategy for experimental design in surrogate model training, based on the Highest Density Region (HDR) of the input parameter distribution. By generating samples uniformly within the HDR, the method significantly improves the accuracy of failure probability estimation compared to conventional random sampling over the full parameter space. Performance is rigorously evaluated using three complementary metrics: leave-one-out error, relative mean squared error (RMSE), and failure probability estimation error. Experimental results across multiple benchmark cases consistently demonstrate that the proposed approach reduces relative RMSE by an average of 32% and lowers failure probability estimation error by one order of magnitude. These improvements are particularly beneficial for tail-probability–sensitive applications, including reliability analysis and Bayesian inverse problems.
📝 Abstract
This paper introduces a practical sampling method for training surrogate models in the context of uncertainty propagation. We propose a heuristic method to uniformly draw samples within highest density regions of the density given by the random vector describing the uncertainty of the model parameters. The resulting experimental design aims to provide a better approximation of the underlying true model compared to the cases where experimental designs have been drawn according to the distribution of the random vector itself. To assess the quality of our approach, three error metrics are considered: The first is the leave-one-out error, the second the relative mean square error and the third is the error generated by the surrogate model when estimating the probability of failure of the system compared to its reference value. The highest density region-based designs are shown to globally outperform the random vector-based designs both in terms of relative mean square error as well as in estimating the probability of failure. The proposed method is applicable within a black-box context and is compatible with existing uncertainty quantification frameworks for low dimensional and moderately correlated inputs. It may thus be useful in case of reliability problems, Bayesian inverse analysis, or whenever the surrogate model is used in a predictor mode.