The real-time wear uncertainty analysis of brake pads is crucial for the braking safety of aircraft. This paper investigates the wear failure of brake pads under thermo-mechanical coupled effects. By considering material degradation and geometric changes of the brake pads during the braking process, a dynamic finite element simulation model is constructed, revealing the dynamic characteristics of brake pad wear degradation. Combining this with a long short-term memory (LSTM) neural network as a surrogate model, the wear amount of brake pads is predicted in real-time. Additionally, the prediction uncertainty caused by sensor error is quantified. Finally, the study monitors a complete braking process of aircraft disc brake pads using a brake pad test bench and conducts simulations and evaluations. The research methodology of this paper, while ensuring accuracy, achieves real-time wear uncertainty analysis of brake pads, providing a theoretical basis for aircraft maintenance and support.