6th International Research/Expert Conference Trends in the Development of Machinery and Associated Technology TMT 2002, Neum, B&H,

18-22 September, 2002

EXPERIMENTAL AND NUMERICAL INVESTIGATION OF THE DROP IMPACT OF FLUID-FILLED CONTAINERS

Aleksandar Karac, Alojz Ivankovic Imperial College of Science, Technology and Medicine London, UK

ABSRACT
Drop impact of fluid-filled flexible containers made of blow moulded high density polyethylene (HDPE) is of considerable concern to containers manufacturers and distribution industries for safety and economical reasons. This is due to potential failure of the containers following the drop impact and subsequent spillage of the transported liquid. The problem falls into a category of strongly coupled fluid-structure interaction (FSI) problems due to comparable stiffnesses of the container and its liquid content. Accurate simulation and successful prediction of containers behaviour must include some kind of liquid-container interaction model coupled with a failure model of the container material. In this work, behaviour of fluid-filled containers under drop impact is investigated using the twosystem FSI model based on the Finite Volume method. Failures of the containers were not considered at this stage. Experimental data are obtained from the drop tests of water filled bottles. A good agreement is found between measured and predicted pressure and strain histories. Keywords: Finite Volume Method (FVM), fluid-structure interaction, drop impact, plastic containers 1. INTRODUCTION In order to measure or evaluate drop impact resistance of a blow-moulded container different approaches can be used: standard testing procedures (e.g. ASTM D2463-95), theoretical predictions (e.g. waterhammer or mass-spring theory[1]), numerical simulation [2,3], etc. The standard procedures provide a critical drop height above which particular container will fail, by using a statistical approach. Thus, containers of different shapes, sizes and material properties must be tested individually, making this approach very expensive in design optimisation, although very quick and useful in controlling the manufacturing process. On the other hand, application of the analytical predictions, i.e. pressure propagation, pressure distribution, etc., is constrained to simple geometry and simple (i.e. linear) material behaviour. Thus, properly validated numerical model is preferable option. 2. EXPERIMENTAL PROCEDURE In order to study the behaviour of drop impact containers and validate developed numerical procedure, a set of experiments is conducted using specially designed rig. The rig, shown in Fig. 1, is an assembly of two aluminium end-caps held together with three steel connection bars, and the specimen with the cylindrical cross section placed in-between. The rig is manufactured to house two types of specimen: bottle-shaped specimen with bottom end removed and bottle with bottom end as originally manufactured. Both specimens are of the same size (diameter intact D=84 mm, thickness t=1.5 mm, height h=160 mm) and type, the first having the bottom end cut off and it is fixed to the bottom-end cap by a worm driven hose clip. There are two main reasons for this approach. Firstly, the influence of the bottom shape and type (rigid and flat as opposed to flexible and complex bottom shape) on the pressure and strain distribution in the bottle can be examined. Secondly, a numerical

simulation of the bottle with flat and rigid bottom can be more easily performed and tested than that of real bottle.
Top end cap Connection bars Water-filled bottle Strain gauges 2 1 PT holder

3 Pressure 1 transducers 2

Worm driven hose clip Bottom-end cap

Figure 1: Experimental rig: left bottle without bottom end; right real bottle The specimen, filled with water levelled at 125 mm, is instrumented with two strain gauges, positioned 25 and 50 mm (or 80 mm) from the bottom, to record deformation of the wall. In addition, three pressure transducers (one positioned axially, and two at the bottom-end cap for bottles without original bottom end) are used to record pressure histories (see Fig. 1). The instrumented rig is dropped onto a concrete floor from a given height. To insure square landing of the rig a set of three U-profile guides is used. Signals from strain gauges and pressure transducers are recorded simultaneously and transferred from an oscilloscope to a computer. 3. TWO-SYSTEM PROCEDURE FOR FSI PROBLEMS The most important part of any FSI analysis is the coupling of the fluid and solid domains, which involves exchanges of appropriate information across the fluid-solid interface. The type of information to be passed across the interface depends mainly on the solution procedure used for each domain. In this work, the two-system procedure, where both solid and fluid domains are solved by FVM, is used. Namely, the FVM has recently emerged as a powerful tool for solving stress analysis problems, and therefore represents a new possibility for solving both fluid and solid domains within a single numerical framework [4]. The procedure was successfully used to simulate wave propagation through the plastic pipes [5], representing a relatively rigid system, as well as for very flexible problems, such as blood flow through arteries [6]. In the present work, the fluid domain is solved first employing a compressible fluid model. Then, pressure from the fluid domain (fluid boundary on the fluid-solid interface) is passed to the corresponding solid boundary on the interface as boundary traction. Using this time-dependent boundary condition, solid domain is then solved. The level of the FSI coupling is governed by the number of iterations used in each time step. In the case where solid structure is relatively rigid, one iteration in each time step can be sufficient in the analysis. This procedure is referred to as explicit coupling. Implicit coupling is a procedure, where a number of iteration is used in each time step to achieve convergence of the solid and fluid domains. 4. MODEL VALIDATION AND DISCUSSION The procedure discussed in the previous section is used to simulate the drop impact test described in Section 2. Due to the axisymmetric nature of the problem, only a part of both domains is considered in the analysis (see Fig.2-left). Two different cases are investigated: (i) the problem with fixed flat bottom end, thus simulating the drop impact of the bottle without bottom end, and (ii) the problem with different bottom-end shape (flat and curved, as shown in Fig.2-right) to investigate the bottomshape effect. In the letter case, a contact procedure is implemented to allow bottle to bounce after the impact. Dimensions of the domains correspond to the actual bottle dimensions (diameter D=84 mm, thickness t=1.5 mm, H=125 mm). The total number of cells for the solid domain is 150 for the first and 162 for the second case, whereas for the fluid domain 1000 and 642, respectively. Material properties for fluid and solid are as follows:

solid (HDPE) modulus of elasticity E = 1.3 GPa; density = 948 kg/m3; Poisson s ratio =0.35 fluid density = 998.2 kg/m3; dynamic viscosity = 0.001 Pas

r v
L Fluid domain

Solid domain Fluidstructure interface Impact surface t

flat flexible bottom end

curved flexible bottom end

=
D/2

Figure 2: Computational domains: left flat rigid bottom end case; right flexible bottom end cases Both domains have the initial speed v=2.8 m/s (corresponding to the drop height of 0.4 m). In the first case the bottom surface for both domains is suddenly fixed. Namely, displacements and velocities at the bottom surface are set to zero, and p/z=0. Solid top surface is assumed to be symmetry plane, whereas the rest of the boundaries are traction free. The pressure at the fluid top surface is atmospheric, simulating the free surface, i.e. no special free-surface model is implemented in the code, with zero gradient boundary condition for velocity. In the second case, the real bottom-end is modelled in order to simulate impact of the bottle with the rigid floor. The fluid stoppage and consequent bounce is the outcome of the simulation. The time step in all simulations is set to 1 s, and total running time is 5 ms.
0.015
0.015
25 mm from the bottom 50 mm from the bottom

0.010

0.010

25 mm from the bottom 50 mm from the bottom

0.005

Strain, strain

Strain, strain

0.005

0.000

0.005

0.000

0.010

0.005
0.015

2 Time, ms

5 Time, ms

10

Figure 3: Strain histories the flat bottom end test: left numerical simulation; right experiment The left part of Fig.3 presents the comparison between strain histories in the bottle obtained by implicit numerical simulation at two different positions 25 and 50 mm from the bottom. As soon as the bottle hits the floor, compressive pressure wave is generated and starts to travel towards the top of the bottle, deforming the bottle wall; firstly, the position nearest to the bottom (25 mm) is influenced by the pressure rise, and then the other positions pressure propagation can be observed by delay between different positions. The process is very similar to the waterhammer phenomena in pipe lines, and is characterised by sinusoid-like wave. The high frequency oscillations superimposed with the main period are due to natural oscillation of the bottle. Similar behaviour, pressure propagation characterised by the delay between different positions, can also be observed in experimental results (see Fig.3 right), which is opposed to the predictions made by Reed et. al. [1]. Superimposed oscillations with higher frequency are also present in the trace. However, the longer low-pressure period can be seen in the experimental results just after 2 ms. This is due to cavitation that takes place during the experiment. In the numerical simulation, cavitation is not modelled, and pressure can drop below absolute zero. This is not a limitation of the procedure. Furthermore, the fracture of the container occurs during the first pressurising period, which is not account for in the simulations.

0.015 rigid fixed bottom end original flat bottom end original curved bottom end 0.010

0.015

flat rigid bottom end original curved bottom end 0.010

0.005

Strain, strain

Strain, strain

0.000

0.005

0.005

0.000
0.010

0.015

2 Time, ms

0.005

5 Time, ms

10

Figure 4: Influence of the bottom-end shape to the strain history in the bottle wall 50 mm from the bottom: left numerical simulation; right experiment Figure 4left shows the influence of the bottom-end shape on the strain history in the bottle wall 50 mm from the bottom end, as obtained from numerical simulations. It can be seen that the loading rate in lower of the bottle with original curved bottom end. This is caused by a gradual stoppage time of the water due to deformation of the bottom end as opposed to a sudden stop when bottle with flat rigid bottom end is used. The shape of the first loading pulse is triangular, and slightly smaller in magnitude. In addition, the following oscillations are much shorter and smaller in magnitude since the bottle bounced. Similar behaviour and good resemblance can be observed in the experimental results, as shown in the Fig.4right. However, the numerical simulation somewhat overestimates the strain magnitude. This is probably due to assumptions of the constant wall thickness and perfect square landing, which is unlikely in reality. 5. CONCLUSIONS This paper presents the combined experimental and numerical investigation of the behaviour of fluidfilled plastic containers subjected to drop impact. Drop impact experiments were conducted on original and modified (bottom-end cut off) bottles. The two-system FSI model with stationary fluid domain boundaries (Euler approach) is used due to its simplicity, stability and efficiency. The results obtained show a good agreement with experimental data, demonstrating the capability of the code to cope with this complex fluid-structure interaction problem. The procedure presents the skeleton of the general, predictive fluid-structure-fracture procedure [7] to be applied to drop impact analysis of fluid-filled containers undergoing failure. 6.
[1] [2] [3]

REFERENCES
P.E. Reed, G. Breedveld, B.C. Lim, Simulation of the Drop Impact for Moulded Thermoplastic Containers, Int. J. Imp. Eng., 24, (2000), 133-153. A. Karac, A. Ivankovic, Behaviour of Fluid-Filled Plastic Containers under Drop Impact, in Proc. Int. Conf. on Computational Engineering Sciences ICES2K, LA, USA (2000) A. Karac, A. Ivankovic, Drop Impact of Fluid-filled Plastic Containers: Finite Volume Method for Coupled Fluid-structure-fracture problems, in Proc. Fifth World Congress on Computational Mechanics, Vienna (2002) I. Demirdzic and S. Muzaferija, Numerical Method for Coupled Fluid Flow, Heat Transfer and Stress Analysis Using Unstructured Moving Meshes with Cells of Arbitrary Topology, Comp. Meth. App. Mech. & Eng., 125 (1995), 235-255. C.J. Greenshields, H.G. Weller, A. Ivankovic, The Finite Volume Method for Coupled Fluid Flow and Stress Analysis, Computer Modeling and Simulation in Engineering, 4, (1999), 213-218. A. Ivankovic, A. Karac, E. Dendrinos, Blood Flow in Deformable Arteries: The Finite Volume Method for Fluid-Structure Interaction Problem, in Proc. 9th ACME Conference on Computational Mechanics in Engineering, Birmingham (2001) in Proc. 10th ACME Conference on Computational Mechanics in Engineering, Swansea (2002)

[4]

[5] [6]

[7] A. Ivankovic, H. Jasak, A. Karac, V. Tropsa, Prediction of Dynamic Fracture in Pressurised Plastic Pipes,