Load-sharing systems are structures where the total load on the system is distributed among the components within the system; some common examples are central processing units of multi-processor computers, cables of suspension bridges, valves in hydraulic systems, kidney system in humans etc. When one or more component within a load-sharing system fails, the surviving components experience extra load that leads to an increase in their probability of failure. The literature on load-sharing systems focuses predominantly on parametric statistical modeling which may sometimes be prohibitive due to the assumptions made regarding the underlying component lifetimes. In this paper, we develop a flexible and data-driven model that is weakly parametric, for analysing load-sharing data. The proposed model does not make strong assumptions on the component lifetimes; instead, it uses piecewise linear functions to approximate the cumulative hazards of the component lifetimes in successive stages of component failure within the system. We discuss various important reliability characteristics under the proposed model. We give examples based on real datasets where the proposed model provides better fit compared to the existing models for load-sharing systems that have been used recently.
Co-author (s): Shilpi Biswas and Ayon Ganguly (Indian Institute of Technology Guwahati)
Journal: Applied Stochastic Models in Business and Industry.
Click here
