Abstract: In tunnel construction with tunnel boring machines (TBMs), accurate knowledge of disc-cutter failure states is crucial to ensuring efficient operation and preventing delays and cost overruns. This study investigates the influence of disc-cutter partial wear on tunneling parameters and proposes a novel method for discriminating partial-wear ratio based on a stacking ensemble model. The time-domain features, including the mean and standard deviation, of torque and thrust are analyzed through a series of scaled-down experimental tests on partial wear. Torque and thrust values increase when a disc cutter becomes trapped and partially worn. The impact of the partial-wear ratio on tunneling parameters appears to be more significant than partial-wear depth. A total of 40 features are selected from the time domain, frequency domain, and time-frequency domain to describe the torque and thrust. The relationships between these features and the partial-wear ratio are analyzed using Pearson coefficients and Copula Entropy. The results reveal that, except for the form factor in the time-domain features, the remaining features exhibit certain linear or non-linear correlations with the partial-wear ratio. Lastly, the proposed model successfully achieves discrimination of the partial-wear ratio and outperforms other commonly used models in terms of overall classification accuracy and differentiation capability in different categories. This research provides effective support for monitoring and health management of disc-cutter failure states.