Evaluating the robustness of classical availability models in non-memoryless failure scenarios

The modelling of reliability and availability in industrial systems has traditionally relied on the assumption of a constant failure rate, as characterized by the exponential distribution. However, in real-world systems, components are often subject to progressive wear, leading to increasing failure rates over time. This paper investigates whether the classical availability formula remains applicable in scenarios where the failure process does not exhibit memoryless property. To this end, the availability behavior of a single device is analyzed through Monte Carlo simulations under different statistical models: the exponential distribution (constant failure rate) and non-memoryless distributions such as the Gaussian and Weibull distributions. The results show that the discrepancy between theoretical and simulated availability remains minimal, confirming the robustness of the classical formula even in the presence of wear-related degradation.