Software-hardware embedded system reliability modeling with failure dependency and masked data

Abstract

Traditional system reliability models often ignore failure dependency between subsystems and the existence of system failure masked data, thereby they can’t accurately reflect the reliability modeling analysis of the whole system. This paper investigates an embedded system that comprises software subsystems and hardware subsystems. In order to more accurately assess the embedded system’s reliability, we proposed a reliability superposition model of the software-hardware system with masked data and failure dependency. In the model, the influence of masked failure data and failure dependency between the hardware subsystem and software subsystem is considered in system reliability evaluation, and the influence of fault dependence is solved by Copula function. Estimating the parameters of this model is a challenging task due to the complexity of the parameters. With regards to this, we proposed an immune particle swarm optimization algorithm with enhanced learning ability (IPSO-ELA), which is used to calculate the parameter’s estimation. Additionally, we investigated the impact of varying degrees of failure dependence on system reliability. Finally, the numerical experiment shown that the proposed model, which considers failure dependency among subsystems, outperforms other reliability models that do not. It can be seen from the experimental fitting results and the reliability trend charts that the system reliability when failure dependence is considered is higher than isn’t considered, which is more realistic.