Accurate segmentation of aortic zones and branches in computed tomography angiography (CTA) is essential for diagnosis and treatment planning of aortic diseases. In this paper, we present a 3D multi-class segmentation approach based on a U-Net architecture enhanced with ResNet blocks in both the encoder and decoder. The model was implemented using the nnU-Net framework with standardized preprocessing, including Z-score normalization, isotropic resampling, and patch-based sampling, alongside data augmentation techniques to improve generalization. We evaluated our method on the AortaSeg24 challenge dataset, which consists of 50 annotated CTA scans. On the final testing set of 40 images, our model achieved an average Dice Similarity Coefficient (DSC) of 0.773 ± 0.028 and an average Normalized Surface Distance (NSD) of 0.805 \(\boldsymbol{\pm }\) 0.038, outperforming the baseline CIS-UNet model (DSC: 0.723 ± 0.058, NSD: 0.746 ± 0.067). Our method performed especially well in segmenting major aortic zones and iliac arteries, while minor label confusion was observed in smaller distal branches. These results demonstrate the effectiveness of using residual connections within the U-Net framework for detailed and anatomically consistent aortic segmentation.