Session-based recommendation (SBR) leverages the interaction sequences of anonymous users within a session to predict their next interaction. Previous SBR methods only leverage a limited number of transition patterns in the session sequence, ignoring the importance of time intervals which can demonstrate user interests and make the sequence more distinguishable. Moreover, they do not effectively address the potential noise present in the sessions. To address the aforementioned challenges, we propose Time Interval Aware Graph Neural Networks for Session-Based Recommendation (TIA-GNN). Specifically, we incorporate time interval information into graph neural network, enabling it to capture more intricate item transition patterns. In addition, we propose leveraging similar sessions to improve the representation learning of the current session while mitigating the influence of noise on it. Comprehensive experiments carried out on three public datasets show that TIA-GNN achieves superior performance compared to SOTA methods.