Underplatform dampers (UPDs) can suppress blade vibration through interface friction dissipation. At present, most of the existing UPD analysis models ignore the uncertainties induced by multi-source factors such as manufacturing, excitation, and wear, resulting in insufficient accuracy of dynamic prediction and optimization of UPDs. Therefore, this paper will deeply consider the uncertain factors such as machining tolerances, assembly tolerances, working load environment, and contact surface wear encountered in the actual manufacturing and working process, and carry out the multi-source uncertainty dynamics research of the UPDs. To solve the "dimension disaster" caused by multi-source uncertainty analysis, an adaptive algorithm based on the non-embedded polynomial chaotic expansion (PCE) is proposed. The algorithm selects the integral accuracy of the best chaotic polynomial by generalizing error convergence criteria and then constructs a surrogate model. This algorithm is in agreement with the calculated results based on Monte Carlo simulation (MCS). Through the quantitative characterization of the above uncertain factors, the efficient prediction of the uncertainty zone of the UPDs dynamic response is realized, the influence law of multi-source uncertainty factors on the vibration attenuation characteristics of UPDs is obtained, and sensitivity analysis of all parameters is carried out, which provides data support for the robust optimization design of UPDs. The simulation results show that the adaptive algorithm based on the uncertainty of multi-source parameters can predict the dynamic response of UPDs structure with high precision and high efficiency, and the uncertainty factors have a significant influence on the nonlinear dynamic characteristics of UPDs, In this case, phenomena such as "resonance band" and "frequency shift" will appear, and the uncertainty of multiple source parameters can cause large fluctuations in the dynamic response of UPDs, especially at the formant of the amplitude response. In addition, compared with other nonlinear regions, the Sobol exponents of uncertain parameters near the formant change significantly.