AI-driven propulsion optimization with a graph neural network-based framework for service level agreement prediction and performance enhancement in aerospace systems

Purpose This paper aims to suggest a graph neural network (GNN)-based framework named GraphSLA for the purpose of enhancing the real-time (RT) service level agreement (SLA) predictions, is suggested in this study. This suggested method aims to optimize the aerospace propulsion systems. Design/methodology/approach The software-defined networking (SDN) method is used by the suggested GraphSLA framework. Here, a wide area network (WAN) configuration with five SDN controllers and one OpenFlow controller are included in the SDN method. Scheduling rules, control algorithms and operating conditions are analyzed by the GNN by identifying the relationships among fuel consumption, thermal dynamics and engine performance. Mean square error, coefficient of determination, mean absolute error (MAE) and mean absolute percentage error (MAPE) were included in the validation metrics. Findings From the outcomes, 98% prediction accuracy with 99.12% high fault detection and 98.43% thermal management efficiency was attained by the suggested GraphSLA. Thus, it is clear from the outcomes that this suggested method have the potential in ensuring RT effective propulsion system. Originality/value This work pioneers applying GNNs for SLA-based optimization in aerospace propulsion systems, offering a transformative approach to autonomous, adaptive control and operational sustainability.