The uncertainty analysis of test results of nonlinear system shows the dispersion of test results. In this paper, an evaluation method of test uncertainty of nonlinear system based on polynomial chaos expansion is suggested. The evaluation method is compared with the method specified in the national standard Monte Carlo Method for Evaluation of Measurement Uncertainty through Ishigami test model. The results show that the suggested method has higher accuracy and efficiency. Taking the seat belt anchorages test as an example, the surrogate model of the test is established by polynomial chaos expansion method, and the uncertainty evaluation of the test is realized. To solve the problem of insufficient sample size, the improved hyperbolic truncation technique and Sobol sensitivity analysis based on polynomial chaos expansion model are used to simplify the model. The results show that the two simplified methods can significantly reduce the demand for samples size. The results show that polynomial chaos expansion and its simplification can be widely applied to the uncertainty analysis for various product tests.