Abstract: The soil arching caused by shield tunnel excavation in deep soft soil layers is directly related to the tunnel depth and creep characteristics of soft soil. This article simulates the entire process of excavation and long-term creep of a deep buried shield tunnel in soft soil through centrifuge experiments, and obtains the temporal variation law of soil pressure values on the excavation surface. This experiment sets up multiple independent model tests to simulate two types of soil overburden thickness conditions: 4D burial depth and 6D burial depth. Based on the performance indicators of the centrifuge and the size limitations of the model, the gravity acceleration is selected as 50g, 70g, and 90g to correspond to the three situations of medium burial depth, deep burial depth, and ultra deep burial depth in actual working conditions. In each experiment, two excavation modes: active unloading and passive loading are adopted. By comparing the effects of different burial depths and excavation methods on soil arch effect, a total of 12 comparative experiments are conducted to analyze the changes in face pressure, soil arch generation, and attenuation laws of shield tunneling. According to the experiment, when excavating in deep soft soil layers (≥4D), unloading excavation mode is used, the long-term soil pressure borne by the shield machine and pipe segments after 2 years of excavation is reduced by about 20% compared to the static soil pressure. Adopting the excavation mode of adding soil pressure, during the long-term creep process in the later stage, the passive soil pressure borne by the tunnel segments gradually decreases, and the long-term soil pressure maintained by the shield tunnel segments increases by about 20% compared to the static soil pressure, which provides important time sensitive reference data for shield tunneling excavation in deep soft soil formations.