Estimation of system reliability in Brownian stress-strength models based on sample paths (Q1260695): Difference between revisions

Let \(\bigl\{X(t),\;t\geq 0\bigr\}\) and \(\bigl\{Y(t),\;t\geq 0\bigr\}\) be two independent Brownian motions representing the magnitudes of system stress and system strength over time. The paper describes the distributional structure of the system first-failure time random variable \(T=\inf\bigl\{t: Y(t)-X(t)\leq 0\bigr\}\) through the study of the associated Markov processes \(U(t)=I\{T>t\}\) and \(V(t)=\bigl\{Y(t)- X(t)\bigr\} U(t)\). The problem of estimating the reliability function \(R(t)=P(T>t)\) using likelihood based methods is also considered. Two types of sampling situations are discussed where one observes either \(U(t)\) or \(V(t)\) at selected time points for \(n\) systems following the stress-strength model. The first situation is related to the estimation of inverse-Gaussian parameters and for the second, the maximum likelihood estimator and its asymptotic properties are described.