Data pre-treatment is crucial for improving data quality in plant data analysis. The researchers developed a new method to detect outliers in a Vacuum Distillation Unit, using a modified 3σ approach for short-term outliers and PCA with Hotelling's T^2 statistics for long-term outliers. The new method showed better accuracy in detecting short-term outliers compared to the traditional 3σ method. Additionally, the PCA method successfully identified long-term outliers, which were further validated by the DBSCAN clustering method. This approach enhances the accuracy of data analysis in industrial settings.