An intelligent simulation result validation method based on graph neural network

With the rapid development of modeling and simulation, there has been a growing concern about the credibility of complex simulations. To validate complex simulation models or systems with dynamic and correlated outputs, an intelligent simulation result validation method based on graph neural network (GNN) is presented. A framework for simulation result validation is proposed, illustrating the process divided into three parts: graph structure modeling for validation data, feature extraction based on graph representation learning (GRL), and bayes factor (BF)-based model's credibility assessment. A graph structure modeling method is introduced to provide a predefined graph structure for subsequent GRL primarily. Next, the interdependencies and dynamic evolutionary patterns among variables are captured by a Multi-level Feature Extraction-based Graph Representation Learning (MFEGRL) model. The similarity of the graph representations is then compared based on the BF to determine the model's credibility. Finally, the effectiveness of this method is demonstrated through a case study focusing on validating simulation models about the terminal guidance stage of a flight vehicle.