This content is not included in your SAE MOBILUS subscription, or you are not logged in.
Prediction of Thoracic Injuries as a Function of Occupant Kinematics
Published January 01, 1979 by National Highway Traffic Safety Administration in United States
Injury-predictive models of the severity of blunt thoracic trauma have been developed using the data gathered in a series of impact tests utilizing eighteen cadaver subjects. Two classes of tests were conducted to generate various levels of injury. The first class of tests was conducted on an impact sled where the subjects interacted with padded or rigid side door structures in side impact. The second class of tests involved controlled energy impacts delivered to the side of the thorax using a flat-faced pendulum impactor. The instrumentation which was designed to monitor kinematics consisted of a matrix of accelerometers surgically placed on the bony exterior structure of the thorax. Two accelerometers were located on the sternum; four were mounted on each side at the most lateral aspect of the fourth and eighth ribs; six were mounted in triaxial clusters at the first and twelfth thoracic vertebrae. The active axes of the accelerometers were chosen to represent the three orthogonal axes, and hence, to provide information on direction of impact. After the tests, autopsies were performed to obtain a level of injury reported as an AIS number.
The injury-predictive models have the AIS number, the number of rib fractures, or derivatives thereof, as the dependent variable and various features of the time-dependent acceleration traces as the independent variables. The features which show the greatest correlation with the level of injury, and hence, which hold the greatest potential for building injury-predictive models have been found to be quantities related to velocities or integrals of acceleration pulses. After quantities such as these were derived from the accelerometer traces, those with the highest correlation to injury measures were selected as candidate independent variables. The predictive models, which were then generated using regression software, are given and discussed.