L. Wang, D. M. Eastwood
ABSTRACT
NOMENCLATURE
B Hardening constant
M Mass matrix
n Hardening exponent
P External applied force
T Current temperature
u Displacement
t Time increment
u Velocity
u Acceleration
TH Homologous temperature
Stress
1. INTRODUCTION
Until now, few papers were published on the topic of numerical simulation of
turbocharger containment. Memhard et al [1] and Zheng Jing [2] have done a
significant amount of work on material testing, aiming to develop the material
plasticity and failure models at various strain rates and temperatures for the turbine
housing materials. Very limited work was presented, however, regarding the
simulation techniques of turbocharger containment. Winter et al [3] did some
pioneering work on the FEA simulation of compressor containment, where relatively
good correlation was shown between FEA and testing. Two of the major drawbacks
of this paper were: firstly no failure criteria were used on the housing material, hence
it may be very challenging to quantify the failure of a housing; secondly, the impeller
was built as three parts to represent the three fragments in testing. This may result
in difficulty of using this FEA model to predict the burst speed and modelling crack
propagation. Most recently, Ramamoorthy et al [4] developed a method to predict
the wheel burst speed and an impressive correlation (2% in speed) to test was
achieved. However, their criterion to verify the housing containment may be limited.
In this paper, the explicit FEA solution method has been chosen to simulate the
containment process, because it is more robust and efficient in modelling problems
that involve high nonlinearities, high speed impact and material fracture. A strain
rate and temperature dependent plasticity material model with fracture modelling
capability is employed to model the turbine housing material behaviour. In addition,
a methodology of optimising the weakening slot design to burst a turbine wheel at
any specified speed is presented. This approach has been implemented in Cummins
Turbo Technologies and good correlation is achieved.
Finite Element solution methods are generally resolved into the implicit method and
the explicit method [5]. The implicit FEA method iterates to find the approximate
static equilibrium at the end of each load increment. It controls the increment by a
convergence criterion throughout the simulation. For a highly non-linear problem, a
large number of iterations have to be carried out before finding the equilibrium. Thus
the global stiffness matrix has to be assembled and inverted many times during the
analysis. Therefore, the computation is extremely expensive and memory
requirements are also very high. It is difficult to predict how long it will take to solve
the problem or even if convergence can be achieved. Thus the implicit method is
preferable to analyse problems under static and simple loading conditions.
The explicit method determines a solution by advancing the kinematic state from one
time increment to the next, without iteration. The explicit solution method uses a
diagonal mass matrix to solve for the accelerations and there are no convergence
checks. Therefore it is more robust and efficient for complicated problems, such as
dynamic events, nonlinear behaviours, and complex contact conditions. However, in
order to obtain accurate results, the time increment has to be extremely small, which
ensures that the acceleration through the time increment is nearly constant.
Therefore it typically requires many thousands of increments. At the beginning of the
time increment (t), based on the dynamic equilibrium equation [6-7]:
P I = Mu (1)
u | (t ) = ( M ) 1 ( P I ) | (t ) (2)
where M is the nodal mass matrix, P is the vector of externally applied force and I is
the vector of internally induced element force. The acceleration of any node is
completely determined by the mass and the net force acting on it. Through time the
accelerations are integrated using the central difference rule, by which the change of
the velocity ( u ) is calculated from equation (3), assuming that the acceleration is
constant:
t (t + t ) + t ( t )
u | (t + t / 2 ) = u | (t t / 2 ) + u | (t ) (3)
2
The velocities are integrated through time and added to the displacement (u) at the
beginning of the increment to calculate the displacements at the end of the increment:
u | ( t + t ) = u | (t ) + t | ( t + t ) u | ( t + t / 2 ) (4)
The element strain increments d is calculated from the strain rate, and then the
stresses are obtained from the constitutive equations:
(t +t ) = f ( (t ) , d ) (5)
In this study, the Explicit FEA solution method provided by commercial FEA software
ANSYS Autodyn has been chosen to analyse the wheel burst and housing
containment of turbochargers.
In the turbine containment testing there are two failure modes of the turbine wheels,
i.e. hub burst and blade detachment [8-9]. To provoke the blade detachment failure,
it is required to cut a weakening slot (either angular or straight) on the backface of
a turbine wheel [9]. However, trial-and-error is sometimes required in designing the
correct weakening slot, to burst the wheel at its target burst speed. Testing inevitably
costs a considerable amount of money and takes time. The objective of this study is
to develop an analysis-led-testing method by optimising the weakening slot design.
An angular velocity of 218750 rpm, which is the target burst speed of this wheel as
per specification of [9], is applied to the wheel only FEA model. Due to the complexity
of the geometry tetrahedral elements (first order, explicit) are used to mesh the
turbine wheel resulting in 1.8 million elements, as shown in Figure 1.
Weakening
slot
Figure 1 Meshing of the weakened turbine wheel FE model
The elastic-plastic properties of Inconel 713C, taken from a uniaxial tensile test, are
used to simulate the material behaviour of the turbine wheel under the high
centrifugal force. A plastic strain based material failure criterion is applied to model
the ductile material failure of the wheel burst. It is essential that the failure criterion
does not over-predict or under-predict the burst speed for a specified weakening slot
design. The failure criterion is achieved by employing the workflow as shown in Figure
2. The comparison between test and simulation is shown in Figure 3.
Table 1 Input factors and levels in RSM analysis of wheel burst FEA
The statistics software Minitab has been used to analyse the effects of inputs to the
outputs. According to the 3D scatterplot and main effects plot, as shown in Figure 5
and 6, high plastic strain (wheel burst) tends to occur if increasing the OD (outer
diameter) or increasing depth of the weakening slot. Furthermore, in this study no
interaction between slot depth and slot OD exists, as shown in Figure 7.
The target burst speed is applied to the weakened wheel as the initial condition of
the FEA model. Impact from wheel fragments to the housing is modelled by using
the penalty formulation of the trajectory contact approach [11], where a contact
event is detected, if the trajectory of a node and a face intersects during one
computation cycle. A local penalty force is calculated to push the node back to the
face. Equal and opposite forces are calculated on the nodes of the face in order to
conserve linear and angular momentum. Frictional effects are considered and a
frictional coefficient of 0.2 is assumed between the wheel and housing.
n
Y = A + Bp 1 + Clnp [1 (TH )m ] (6)
where p is the effective plastic strain, p is the plastic strain rate; TH is the
homologous temperature, which is defined by
TTroom
TH = T (7)
melt Troom
There are five material constants in the Johnson Cook model: A is the initial yield
stress,B is the hardening constant, n is the hardening exponent, C is the strain rate
constant, and is thermal softening exponent.
To simulate the material failure there are a number of material fracture modes [13-
15], most of which, however, require comprehensive testing. In this study, a
relatively simple plastic strain based criterion is used to model the housing material
failure and good correlation is shown between FEA and testing. Element erosion [6]
is employed to simulate the penetration of the housing due to the high speed impact
from the wheel fragments, as shown in Figure 10. It is clear that the blades detach
from the wheel at the region near the weakening slot, gradually impacting the
housing. The wheel fragments reach speed zero at the end of the impact and the
housing successfully contains the wheel fragments in this case study.
(a) (b)
(c) (d)
(e) (f)
Figure 10 Wheel velocity plots of the housing containment FEA model
6. COMPARISON BETWEEN TEST AND SIMUALTION
The explicit FEA method of modelling the wheel burst and housing containment is
applied to other case studies, where good correlation between test and simulation
has been achieved. Figure 11 compares the wheel burst FEA with test by using the
straight cut weakening slots. Again it demonstrates the importance of the weakening
slot design. In the case study 3 shown in Figure 11(a), all three blades detach from
the hub. Nevertheless, by using the slot design of case study 4, as shown in Figure
11(b), the middle blade remains intact, indicating that the slot depth may not be
deep enough to result in the detachment of all three blades. Figure 12(a) and Figure
12(b) illustrate the correlation in terms of the damage at the inner wall and inlet of
two turbine housings, respectively. Furthermore, the technique of turbine housing
containment FEA has been applied to the compressor containment with Aluminum
impeller and ductile iron housing. Figure 13 showcases the correlation of the fractured
impeller from the compressor containment test and FEA simulation.
In this paper two major aspects of the turbine containment, i.e. wheel burst and
housing containment, have been simulated by explicit FEA; which demonstrates
excellent correlation with testing. Based on this study, the following conclusions may
be drawn:
a) Wheel burst tends to occur if increasing the outer diameter or depth of the
weakening slot.
c) The Johnson Cook strength model is suitable to simulate the plastic behaviour of
turbine housing, which is subjected to large strain, high strain rate and high
temperature in the containment simulation.
d) A plastic strain based failure criterion combined with element erosion technique
may be capable of modelling the ductile material failures in turbo containment FEA.
ACKNOWLEDGEMENTS
The authors would like to thank Cummins Turbo Technologies to allow publishing this
paper. Our sincere thanks also go to Mr. Owen A Ryder, Mr. Jeffrey D Jones and Mr.
Simon J Tooley who provided many valuable suggestions and proof read this paper.
We would like to acknowledge the consistent support from other team members of
the Applied Mechanics department.
REFERENCE LIST