A method of computing dynamic roof crush in rollover accidents has been proposed (Bidez, et al., 2005; Cochran et al., 2005). The method used data obtained from accelerometers mounted to the roof rails of sport utility vehicles, along with other measurements, to compute the instantaneous deformation of the roof rails during dolly rollover crash tests.
We examined the feasibility and practicality of this methodology in three ways. First, the theoretical derivation was examined. Errors appeared to have been made in deriving and/or interpreting the equations used to compute instantaneous roof crush. Next, a three-dimensional dynamic rollover simulation program was run to produce ideal acceleration data (Yamaguchi et al., 2006, 2005). Using these data, the equations in original, uncorrected form predicted dynamic roof deformations when none existed. When the equations were corrected, the simulation data yielded proper roof positions and no roof deformations. Finally, a dolly rollover test was conducted to generate real-world accelerometer data. The vehicle was heavily braced internally to avoid producing significant roof crush. Vibrational “noise” present in the real-world accelerometer data caused a large amount of calculational error in the dynamic roof crush calculations. Sensor positioning and angular alignment errors also may have contributed. These results indicate that the method of computing dynamic roof crush from accelerometer data is theoretically feasible when correct equations and ideal data are used, but is impractical using measured real world accelerometer and angular rate sensor data.