Abstract: Aiming at the characteristic of complex polymorphism of the pantograph system and the difficulty in expressing the accuracy of the failure probability, and the current situation that the existing research is mainly stuck in the fuzzy Bayesian network combined with the evidence theory, the hyper-ellipsoidal Bayesian network is introduced for the first time into the reliability analysis of pantograph, which avoids the extreme value of the fuzzy Bayesian network interval and further defines the probability value interval of the root node. Then the hyper-ellipsoidal Bayesian network is used to solve the reliability parameters such as leaf node failure rate, sensitivity and posterior probability of the pantograph system under different fault states, and the high-risk events of the system are found out. Compared with the fuzzy Bayesian network, it can be seen that the interval of the hyper-ellipsoidal Bayesian network is smaller, which verifies the correctness and practicability of the new method.

Key words: pantograph, multi-state, hyper-ellipsoidal Bayesian networks, sensitivity, posteriori probability