Knowledge Graphs (KGs) have by now become a form of knowledge representation and they are the cornerstone of several industrial applications. The ever-increasing interest in this technology is due to its underlying abstract structure, which effectively facilitates domain conceptualization and data management. As a result, KG is the main driver of several Artificial Intelligence applications. Additionally, we live in an interconnected world that empowers KGs. Graph theory provides a way to model and analyzes such interconnectivity. In graph theory, graphs are generally given in advance or easily abstracted from practical problems. However, for many real-world scenarios, the graph is uncertain. As a result, one needs to design or learn graphs from data before any analysis. Therefore, constructing a high quality graph has become an increasingly relevant research problem, which inspired many graph construction methods in the past years. The simplest scenario of graph connectivity is when the graph associated with a problem is physically well-defined. Examples of such graphs are manifold, including electric circuits, power networks, linear heat transfer, social and computer networks, and spring-mass systems. We are interested in crashworthiness studies in vehicle development that, converting a crash simulation to a physical graph is challenging and an unexplored field of research. However, we believe that such a representation enables comparing simulations, emphasizing unexplored design of experiments, detecting load-paths, and correlating different designs to improve design guidelines and support recommendation systems. Consequently, converting a finite element (FE) crash simulation to a physical graph is the focus of this research. The physical graph of crash simulation is the extension of our earlier investigation research in. There we looked at different segments of CAE data with crashworthiness use-case. The earlier graph modeling contains features extracted from simulation results at the Part level; however, the connectivity of the parts needed to be included. Here we add edges between Part nodes of each simulation to represent the energy flow during the crash between parts. The parts connectivity requires extracting features from the structure of the vehicle. For this, we will consider parts as boxes to simplify the problem, group them as components, and later shape the graph’s structure from the boxes. We will focus on load-path detection among different applications of physical graph representation. In the crash simulation, the load-path is the sequence of the parts that absorb the most proportion of the energy caused by the impact. Accordingly, we extend the graph database modeling to group the parts as components and compare the load-path. Besides, we develop a method to detect components and load-paths for each simulation automatically. The nodes’ energy features and the graph structure are the input to the graph analytics to embed the load-path for each simulation. Besides, the grouping enables grouped feature extractions, for example, energy features introduced in. The ultimate highlight of this work is showing the usefulness of graph modeling for CAE. Here the investigation is implemented on a full-front load-case; however, corresponding studies can be undertaken for different impact directions and load-case scenarios.