In this paper we describe a preliminary version of a frontal impact crash sensing algorithm capable of continuously predicting the severity of a crash in real time. This kind of algorithms could be used to control an airbag system with a variable output inflator, which supplies a variable amount of gas into the airbag on demand.

The algorithm consists of two parts linked in series. The first part categorizes the class of an event. The second part predicts the severity of the crash using a function of the occupant free flight displacement and time.

Linear regression and neural network analyses were performed separately to determine the coefficients for the severity function of each crash mode. The algorithm was implemented in Simulink and validated with test data. While both analyses achieved reasonably good correlation between the severity of each event and its corresponding severity function, the neural network analysis generally provided a better correlation.