duration of the oscillations observed. The delay in signals between
the two positions is not as pronounced as when the bottle with
a flat and rigid base is tested, since the distance between the
positions is shorter. The shape is more triangular or sinusoidal than
trapezoidal, which could support the dynamic character of the
event, as in Ref. [7] (mass-spring model). However, this can also be
attributed to a shorter valve-closure time, as explained earlier,
causing the tendency to a triangular shape (waterhammer).
When the trace is compared to that of the bottle with flat and
rigid base (Fig. 9), it can be seen that the pressure rise rate is
significantly affected; a strain rate of 4 s1 is measured. The
maximum pressure is around 40% lower than that measured when
the base is flat. Two separate factors can be responsible: (i) the
effect of base deformation and therefore slower water arrest on
impact causes the tendency to a triangular shape; it has to be
mentioned that the base of this container type has a very complex
curved shape of a slightly bigger diameter than the rest of the
bottle, and (ii) the imperfect impact between the rig and the floor,
and the bottle and the rig; this double impact may prevent a square
landing.
Tests with real bottles show that the base can have significant
effects on the behaviour of containers subjected to drop impact not
only on the pressure (stress) magnitude, but also on the shape of
the histories recorded. Different base shapes affect the results in
differentways, so the tests with a flat, rigid base are used to validate
numerical procedures.
4.2. Numerical simulations
Fig. 10-left shows simulated pressure histories from three
monitoring positions, which correspond to the positions of a pressure
transducer (PT) and two strain gauges (SG). The PT history is
related to the pressure history from the middle base position. SG
histories are obtained from two positions along the bottle wall: 25
and 80 mm from the base. It is clear that all traces follow a similar
pattern, the high frequencies being dominant only in the PT history
in the first 1 ms.
The effect of smoothing a simulated PT history by averaging over
100 points is shown in Fig. 10-right. It can be seen that the highfrequency
oscillations are still present in smoothed trace, and some
low-magnitude oscillations with a short period also remain. The
latter have period of 0.9 ms and are related to natural oscillations of
the bottle, as explained in the experimental part. On the other
hand, the origin of the high-frequency oscillations is not that clear.
It was previously believed that these frequencies were of numerical
nature caused by interactions on the solid-fluid interface (Greenshields
et al. [16]). However, detailed analysis of the pressure traces
and pressure distribution in the fluid domain shows that they have
a physical meaning, and are related to the presence of unconfined
waves in the fluid, as explained below.
Fig.11 shows the pressure distribution in the fluid domain in the
first 0.2 ms, with frame rate of 10 ms. The colours do not have
the same values in the frames, since the figure is only used to
clarify the origins of the high-frequency oscillations in the pressure
traces. The graph (at the lower right-hand corner) represents the
pressure history at the middle of the base (lower left-hand corner of
each frame). At the instant of impact, the fluid away from the wall
has no information about wall movement and a wave starts travelling
at the unconfined speed of 1485 m/s ð
ffiffiffiffiffiffiffiffiffi
K=r
p
Þ, producing
a pressure magnitude of 41.6 bar ðr$cun$DvÞ. This high pressure
region is clearly visible in the middle part of the bottle base (frames
2 and 3). On the other hand, fluid particles next to the bottle wall
start moving radially outward due to deformation of the wall. This
fluid movement radiates a radial decompression wave towards the
bottle axis. After approximately 30 ms (D/2cun ¼28 ms) the wave
reaches the axis (p