Multivariate tests for comparing lifetimes of parallel systems

Life testing of engineering systems with dependent components requires robust multivariate methods. Parametric approaches depend on restrictive assumptions, limiting their use in complex or unknown lifetime distribution settings. This study evaluates nonparametric methods for comparing parallel system lifetimes under minimal sample requirements. Three distribution-free tests are considered: the rank-energy test, the Wilcoxon-type rank-sum precedence test and the Lepage test, the latter two applied to Minkowski distances from lifetime vectors. Results suggest that the Lepage test on Minkowski distances generally outperforms the other two, offering a robust method to compare multi-component system lifetimes.