Remaining Useful Life prediction has long been a critical task in the field of industrial equipment maintenance. As equipment complexity continue to evolve, traditional prediction methods based on mathematical and physical modeling often fail to deliver satisfactory results. Data-driven approaches, tend to overlook the spatial relationships between features. Graph Convolutional Networks have gained attention for their ability to capture spatial feature dependencies across nodes, making them particularly suitable for RUL prediction tasks. However, while considerable focus has been placed on spatial feature learning, the sensor relationships between graph nodes are often either overlooked or inadequately represented. We propose a prediction method called Sensor-Spatial Fusion Graph Network (SSF-GCN). This method integrates sensor channel features, obtained through convolutional layers, with spatial features from the graph structure using an attention mechanism. Additionally, leveraging the Message Passing Neural Network (MPNN) framework, we employ the attention mechanism to efficiently aggregate feature information from distant neighboring nodes. The effectiveness of SSF-GCN is demonstrated through experiments conducted on the publicly available C-MAPSS Dataset. The experimental results show a significant improvement in predictive performance over SOTA methods.