A radial sampling-based subregion partition method for dendrite network-based reliability analysis

In sampling-based reliability analysis, a constraint with multiple failure points may lead to an inefficient iteration process and inaccurate results. To examine this problem, a novel analysis method is proposed in this article, which achieves multiple failure point-based constraint model construction. In the proposed method, a radial sampling-based subregion partition method is presented to locate the potential failure subregions that may have a failure point and to construct a subregion partition model to find the model refinement points in parallel. In addition, a new machine learning algorithm, the dendrite network, is adopted to construct the constraint model and the subregion partition model, and a network-matched learning function is designed to assist dendrite network-based model refinement. Test results demonstrate that the number of training samples is decreased compared with other citation methods.