Reliability evaluation of the drive system of a CNC swivel head with Bayesian networks

The computer numerical control swivel head drive system has a number of states from ideal operation to complete failure, and dividing these states into success and failure is not rational and may even lead to fatal errors. Given this problem, this study proposed a method of building a multistate reliability evaluation model based on a Bayesian network (BN). First, BN was built in accordance with the mechanism of the drive system and in consideration of the interaction between different failure modes. Second, stable working data were used to determine the failure probabilities in the different life cycles of nodes. Last, a probability distribution table was employed to describe the multistate characteristics of nodes, and a BN model of the multistate system was built. The ranking of the importance of risks that may lead to drive system failure was determined using the posterior probability calculation method of BN.