Die-casting is a typical multidisciplinary system, in which there are many disciplines involved such as hydrodynamics, heat transfer and elastic-plastic mechanics and their coupled relations are intricate. Moreover, there are inherently epistemic uncertainties in die-casting. This paper aims at reliability modeling and optimization of diecasting. Firstly, finite element analysis is applied to simulate the process of die-casting, and the quadratic response surface is used to represent the multidisciplinary optimization model of die-casting, in which the evidence theory is used to represent the epistemic uncertainty. Then, a reliability-based multidisciplinary optimization (RBMDO) procedure for diecasting is proposed. The results show the RBMDO procedure possesses high computational efficiency in optimizing multidisciplinary system with epistemic uncertainty.