Abstract: Common cause failure, a kind of dependent failure mode in many systems, is initiated by some common causes. Study of common cause failures is confined to binary systems presently. In the light of non-repairable multi-state series-parallel systems with common cause failure, the mechanism of establishment and effect of common cause failure is represented, and a new reliability assesment model is put forward based on load discretization method and probability generating function approach. The new model can manifest the actions of partial failures in some components with a common failure cause and the corresponding accumulative effects on entire system performance. It is shown that the common cause can inevitably induce failure-dependence of all components in the common cause group. If some component-failures are dependent statistically due to a common cause, probability all components in the same state increments, which can effects the reliability assessment. The load discretization method realizes the approximate decoupling for failure-dependency, and the probability generating function approach is proved to be very convenient for different types of algebraic combination owing to relatively small computational resource costs. The model provides a new path for reliability study of multi-state systems with common cause failures.

Key words: Common cause failure, Discretization, Generating function, Multi-state system, Reliability