This paper was produced for the 2019 NAFEMS World Congress in Quebec Canada
Reasonably accurate simulation of airbag deployment is an important part of a crash analysis. The different sizes, shapes, initial configurations and deployment methods for airbags demand a robust simulation approach. In almost all cases, the airbag is tightly folded, rolled-up or crumpled-up in a small space. When the inflator is fired, the fast moving gases push through the layers of airbag folds during the deployment phase. In some cases, a small amount of gas is allowed to escape through vents in the airbag to enhance the cushioning effect. The airbag deployment is a rapid event and typically lasts between 20 to 30 milliseconds. Explicit integration finite element analysis is therefore suitable for such simulations. The key aspects in an airbag deployment simulation are resolving contact between the folds of the airbag, resolving contact between the airbag and other objects, and the accurate modelling of the gases and their interaction with the airbag. This paper focuses on the later aspect, namely the modelling of gases and their interaction with the airbag.
Previously existing methods for modelling gases in Abaqus/Explicit have achieved varying degrees of success. In some cases these methods either fail to accurately capture the non-uniformity of pressure during the deployment process (the so-called UPM methods), or have problems in resolving the flow of gas through narrow spaces (such as Coupled Eulerian Lagrangian approaches). The lumped kinetic molecular method is a particle-based method. It derives from the kinetic gas theory, which views gases as a collection of randomly moving molecules that collide elastically with each other as well as with the surrounding walls. A real gas consists of an enormous number of molecules of extremely small size. This renders modelling of a real gas next to impossible. In the lumped kinetic molecular method, we assume that the numerical gas consists of particles that are orders of magnitude larger than gas molecules. Further, orders of magnitude fewer particles are required to model a given amount of gas. Despite the smaller number of larger size particles, it is assumed that the particles behave like gas molecules, i.e. they move around randomly and collide elastically with each other as well as with the surrounding walls. These assumptions make the numerical treatment of gas feasible in a thermodynamically consistent fashion. The paper first presents simple tank test results to validate the behaviour of the particles against analytical results. Next, examples of real airbags are presented. These results are compared with experimental results where available.