Design and Analysis of Shrink and Press Fit Joints

Morrisey, P., Eischen, J., Dow, T.

Precision Engineering Center
North Carolina State University

Introduction

Pulsed Power experiments under way at Los Alamos National Labs use short pulses (<5 µsec) of
high current (>107 amps) to create a large magnetic force on the cylindrical “liner” carrying this
current. As a result of this force, the liner implodes onto a target positioned at its centerline.
Creating a large force on the target is the goal of the experiments. The liner is a thin-walled
aluminum cylinder (thickness of 1 mm with a radius of 50 mm) captured at the ends by two
copper guide planes using either press or shrink fits. The contact between the liner and glide
planes is critical to the success of the experiment. The analytical challenge for investigating
these components is predicting the presence of a partial shrink fit; that is, a shrink fit whose
length is less than the apparent length of the interface between mating cylinders. Such stress and
deformation distributions are difficult to calculate using analytical shrink fit theory because of
the varying radial deflection profile along the length of the liner. A second goal is to design an
initial non-cylindrical liner such that when it is assembled onto the glide planes, it will assume a
cylindrical shape. This cylindrical geometry will lead to uniform implosions of the liner and are
deemed to be highly beneficial for the success of the experiments.

A model of the liner/glide plane interface was created using the finite element code, LS-DYNA,
distributed by Livermore Software Technology. This code is a general-purpose, transient
dynamics finite element program originally developed for simulations of projectile penetration,
blast response and explosives. It has the ability to model inelastic material behavior and to treat
contact constraints. The FE model was used to address the following: mechanics of the
interference fits, physical description of the contact surfaces between the liner and glide planes,
joint void (gap) description, shape of the assembled liner and material property effects. Different
shrink fit geometries were investigated to evaluate their effect on the assembled shape, joint
interface gaps and low interface stress magnitudes. The results of the finite element models were
also compared to a scale-model of the liner/glide plane system built at the PEC. The radial
deflection of the diamond turned thin-walled liners fitted to glide planes were measured using an
optical interferometer as well as a contact profilometer. These measurements verified the FEM
results and the technique proved to be a tremendous aid in accurately predicting the shape,
deflection, and stress distribution of such composite cylindrical elements.

Model and Results

Copper glide planes are inserted in either end of the cylindrical liner, typically via a thermal
shrink fit process, though more recent designs have used a modified geometry with a mechanical
press fit process. These glide planes serve as a mechanism to transfer tremendous electrical
currents to the liner, which leads to a cylindrical liner implosion. To optimize the results of these
experiments, it is necessary for this cylindrical implosion to occur very uniformly. Therefore,
manufacturing dimensionally precise liners with minimal surface flaws is essential.
Unfortunately, the shrink fit process naturally results in residual stress and distortion in the liner
walls, which need to be compensated. A schematic illustrating the assembly orientation of a
single wall liner with glide planes is shown in Figure 1. The nominal diameter of the liner is
100mm, the height is 55mm, the wall thickness is 1.2953mm, and the nominal interference fit is
15.24µm (0.0006in).

Glide Planes (copper)

Liner (1100
aluminum)
Joint

1.3mm

100mm

Figure 1- Cross section of liner and glide planes (NTLX Shiva Star/Atlas
Configuration)

Figure 2 shows the baseline mesh modeling the geometry of this axi-symmetric system. The axi-
symmetric modeling does ignore some details of the glide planes, such as holes and ports, but
this is not expected to impact the results of interest in any significant way. Each component is
modeled in LS-DYNA as an elastic-plastic material, thus any yielding that could occur is
considered. The liner is composed of 1638 elements with 8 elements through the thickness, and
the large and small glide planes are composed of approximately 2000 elements each. For this
preliminary FEA model in particular, a composite liner is not used. A uniform temperature
change of -100°C was imposed on the glide planes during simulation. The parts were then
slipped together, followed by a return to ambient temperature with contact modeling between the
parts activated during the transition to equilibrium. The glide planes expanded during this phase,
deforming the liner as expected.

A displacement vector plot for nodes within the liner is shown below in Figure 3a and 3b. The
lengths of the arrows indicate magnitude of radial displacement in the liner wall, with the peak
deflections occurring at either end of the liner. Note that the vectors are magnified for an
enhanced visual representation of the displacement behavior. The displacement decays with
distance from the liner ends and actually becomes negative, indicating inward radial movement
at the mid-length position of the liner.
Figure 4 shows data along the longitudinal axis (3 data scans 120 deg. apart) on the OD of a
scale-model liner/glide test specimen in comparison to the LS-DYNA liner deflection profile.
This data was obtained with the TalySurf Profilometer. Excellent agreement is apparent in this
comparison, providing a validation of the LS-DYNA contact modeling capability.

Figure 2- LS-DYNA mesh used for LANL prototype simulation

Figure 3a- Magnified vector plot of radial displacement along the liner
 Figure 3b- Zoomed view of the radial displacement vector plot

Figure 4- Comparison of LS-DYNA and scale-model test specimen liner displacement values for
three longitudinal scans