Update: June 1998: A common cause of single vehicle rollover crashes - the driver "overcorrects" after veering off the edge of the tar. See K.W. Ogden, The effects of paved shoulders on accidents on rural highways, Accident Analysis & Prevention 29 (3) (1997) pp. 353-362.
There were a total of 129 incidents where vehicle at least rolled onto roof.
The average time per roll for corkscrew and side-on crashes was very similar. A typical single full roll took 2.3 seconds. A typical double roll took 1.5 seconds per roll and a typical multiple roll ( 3 or more) took 1.1 seconds per roll. The average for crashes involving one or more rolls was 1.7 seconds.
Slow motion analysis of the videos revealed that substantial changes in the angular velocity occur as parts of the vehicle contact the ground. This results in high tangential forces on the occupants. In some video frames the head and arms of occupants can be clearly seen extended well outside open/broken side windows. This appears to be a whipping action which, fortunately, tends pull to occupant back inside the vehicle just before the adjacent roof edge makes contact with the ground.
Case 1: Occupant's head well outside window during the second complete roll. His head whips back inside the vehicle just before it hits the ground. Note the occupant is wearing a harness seat belt and helmet. He looked unharmed after the incident!
This animated GIF might work on your browser:
Case 2: Occupant's helmeted head breaks the side window just before that side of the vehicle contacts the ground. Again the occupant was wearing a harness seat belt and was apparently unhurt - but probably had a sore head!
Key features of this analysis are:
As the vehicle rolls over a "corner" the C of G reaches its highest point and, if the speed of the roll is sufficient, the acceleration of the C of G might go positive (note that the effect of Earth's gravity has been included in the above values). This indicates that the vehicle loses contact with the ground. In some of the rally crashes this effect was so strong that the roof did not contact the ground at all during the first half roll.
The highest vertical acceleration occurs when the underside of the vehicle is in contact with the ground, at the start of the roll and at the end of a full roll. This is a manifestation of the location of the C of G of the vehicle, which is closer to the underside of the vehicle than the roof. In typical passenger cars the vertical distance from the C of G to the roof is similar to the transverse distance from the C of G to the side of the vehicle and therefore there is relatively little vertical motion of the C of G as the vehicle rolls from its side to the roof to the other side. Low aspect ratio vehicles such as sports cars have a smaller distance between the C of G and the roof and therefore the vertical loads occuring when the roof is in contact with the ground can expected to be higher. This might partly explain the finding by Moffatt (AAAM 1995) that high roof vehicles have generally less roof damage than low roof sports cars.
The change in vertical velocity during each quarter turn is estimated about 2m/s for this scenario.
Note that during the initial tripping event it is likely that there will be substantial horizontal deceleration which will tend to throw the occupants in the direction of the roll. This appears to be the mechanism of ejection for the unrestrained dummy in the paper by Habberstad et al (1986) - the dummy was ejected before the vehicle reached the first quarter of the roll.
After the tripping event which initiated the roll the vehicle will be moving sideways with a typical horizontal velocity of 11 m/s (assuming the trip speed was just sufficient to cause the roll - see Gillespie, 1992 p326) therefore the first two contacts of the roof with the ground will probably involve relative speeds of about 6 m/s (11 - 5) and the impact will tend to increase the speed of rotation of the vehicle. In effect, the occupants will experience angular accelerations in a direction opposite to the direction of the roll during these first two roof contacts (and opposite to the direction in which they were thrown at the start of the roll).
As the horizontal speed of the vehicle drops (due to the braking effect of the ground impacts) and the rotational speed increases, the tangential speed of the corner of the roof may eventually exceed the horizontal speed of the vehicle and the impact will tend to decrease the speed of rotation of the vehicle. In effect, the occupants will then experience an angular acceleration in the same direction as the roll. The observation of rally crashes where the occupants head and arms are extended outside the side window are probably due to this angular deceleration, the peak of which would usually occur as the vehicle tips over on its wheels near the end of the first roll or at the start of the second roll. At each of these points the C of G of the vehicle is passing through its highest point therefore the occupants tend to become "weightless". An unrestrained occupant has a high risk of being ejected at this point, if the side window is open or broken.
Rotational speed will have a similar step-like profile to horizontal velocity. To obtain a rough estimate of the change in rotational speed during each ground contact, assume that the maximum rotational speed is reached after half a revolution (at the third ground contact). For a maximum rotational speed of 6 rad/s (based on a average of 3.8 rad/s for a full roll) the change during each contact will therefore be about 2 rad/s. This is equivalent to a linear velocity change of about 2m/s in the region of an occupant's head, in a direction tangential to the C of G of the vehicle.
Acknowledgements: Thanks to the Rallysport Promotions (Rally Hits Volume 2) and Duke Videos (Crash Kings) for the fascinating videos. The Federal Office of Road Safety sponsored the research project on which these notes are based. Dr Michael Henderson was the principal consultant on that project. Details of references are contained in the research report. The report was released in July 1999 as CR 176 - see FORS.
This web page resulted from my playing with a new toy - the Snappy
Video Snapshot - which makes it easy to capture stills off video.
resolution of the pictures on this page was intentionally low so that
files sizes were manageable. For a clearer picture buy the videos!