Taking the 3UPS/S parallel manipulator as the research object, a output error prediction model of parallel manipulator based on virtual experiment and BP neural network was constructed, which could predict the output error of parallel manipulator. A virtual prototype model with the input error was established by considering the hinge installation error with the hinge axis. It was assume that the error of the components in the mass production was in the normal distribution, and the input error samples of the multi group followed the normal distribution were constructed, and then the BP neural network prediction model of the parallel manipulator was established. The results show that the BP neural network model can predict accurately output pose error of the manipulator, which provides a new basis for the error analysis and accuracy synthesis of parallel manipulator.