Abstract: For low efficiency and poor accuracy problems in reliability analysis of flexible mechanism, support vector machine regression extremum method (SREM) is proposed in dynamic reliability analysis of flexible mechanism. A small amount of samples for flexible mechanism dynamic response extremum are generated by Monte Carlo (MC). Based on these samples and support vector machine regression theory, the surrogate model of flexible mechanism dynamic response extremum is established. Dynamic response reliability of flexible mechanism can be calculated out by the surrogate model. In crank-rocker flexible mechanism, the reliability of flexible link deformation and the reliability of rocker swing angle are assessed by SREM, MC, artificial neural network surrogate and quadratic polynomial surrogate respectively. By comparing calculation results of four methods，SREM is proved to be high precision and high efficiency in dynamic reliability analysis of flexible mechanism, and the feasibility and application of SREM is verified in dynamic reliability analysis of flexible mechanism.

Key words: flexible mechanism;response surface;support vector machine;dynamic reliability;regression extremum method