The Split Hopkinson Pressure Bar (SHPB) experiment is commonly employed to assess the dynamic mechanical properties of materials under high strain-rate conditions (10²-10⁴ s-¹) through the propagation of elastic stress waves via pressure bars. The precision and dependability of SHPB measurements are heavily influenced by the alignment of the specimen with the bars. Misalignment can lead to flexural vibrations, causing waveform distortion and undermining the assumption of onedimensional stress waves. While previous research has explored the impact of misalignment on waveform characteristics, pinpointing the specific sources of distortion from measured signals remains a challenge. This study introduces a machine learning-based classification method that extracts features from distorted SHPB waveforms to identify the type of misalignment. Incident wave signals under various misalignment scenarios were simulated using the commercial finite element software LS-DYNA, and the extracted features were utilized to create a training dataset. Several machine learning models, including XGBoost, were trained and evaluated, with XGBoost yielding the highest accuracy and F1-score. The trained model was then applied to experimentally measured distorted waveforms to validate its effectiveness. This proposed approach facilitates the automated diagnosis of distortion sources in SHPB data, reducing the need for manual interpretation and improving analysis efficiency.