Professional Documents
Culture Documents
org
Abstract
Lubrication in automotive gear box address the balance between the reliability
issue and the CO2 emission saving; and depends mainly on mass of oil and
internal oil pathways. The efficiency of the lubrication is a balance between the
oil mass usually limited as low as possible and the split of the oil mass in the
different parts of the gear box. The objective is the maximum of the oil mass
should be used for lubrication. The partway depends on both speed shafts
conditions and internal geometric design of the gear box, including casing,
gears and bearings. Consequently, geometric scales vary from 1 to 100.
A preliminary study was done on a simplified test case and first conclusions
were applied in balance to CPU time consuming. Results were compared to
experimental data and a second study was done using local refinement of the
size of SPH particles.
Presented at the NAFEMS World Congress 2019 Québec City, Canada | 17-20th June 2019
© NAFEMS 2019 REPRODUCTION AND REDISTRIBUTION PROHIBITED nafems.org
1. Introduction
The areas of interest of the SPH method are the dynamic flow and/or flow with
complex fluid or solid interfaces. These areas often coincide with the
limitations of more traditional meshed methods. The SPH method is not in
competition with conventional methods but complementary. In automotive
applications, lubrication problems are often dynamic flows with complex
interfaces. The SPH method can therefore respond to questions that are major
issues in the design of gearboxes. This paper will present validations of
lubrication simulation in automotive gearbox with SPH method.
2. Governing equations
𝐷𝐷ρ (1)
= −ρ∇. 𝑢𝑢
�⃗,
𝐷𝐷𝐷𝐷
𝐷𝐷𝑢𝑢
�⃗ 1 (2)
= − ∇P + υ∇2 𝑢𝑢
�⃗ + 𝑔𝑔⃗,
𝐷𝐷𝐷𝐷 𝜌𝜌
𝐷𝐷𝑥𝑥⃗ (3)
= 𝑢𝑢
�⃗,
𝐷𝐷𝐷𝐷
where the density, velocity, pressure, kinematic viscosity, gravity and position
are respectively noted 𝜌𝜌, 𝑢𝑢
�⃗, 𝑃𝑃, 𝜐𝜐, 𝑔𝑔⃗ and 𝑥𝑥⃗.
In order to close this system, the following equation of state is used throughout
this paper:
Presented at the NAFEMS World Congress 2019 Québec City, Canada | 17-20th June 2019
© NAFEMS 2019 REPRODUCTION AND REDISTRIBUTION PROHIBITED nafems.org
where 𝛾𝛾, 𝜌𝜌0 and 𝑐𝑐0 stand respectively for polytropic constant, reference density
of the fluid and nominal speed of sound. Here 𝛾𝛾 is equal to 7 for the simulation
of liquids (Tait, 1888). In order to obtain a quasi-incompressible flow, 𝑐𝑐0
should be chosen as 𝑐𝑐0 ≥ 10𝑢𝑢𝑚𝑚𝑚𝑚𝑚𝑚 where 𝑢𝑢𝑚𝑚𝑚𝑚𝑚𝑚 is the maximum expected
velocity throughout the whole simulation. This ensures a Mach number
respecting 𝑀𝑀𝑀𝑀 ≤ 0.1.
3. SPH background
In its original form, SPH uses a pressure gradient and velocity divergence
defined as:
Presented at the NAFEMS World Congress 2019 Québec City, Canada | 17-20th June 2019
© NAFEMS 2019 REPRODUCTION AND REDISTRIBUTION PROHIBITED nafems.org
〈∇. 𝑢𝑢
�⃗〉𝑖𝑖 = � 𝜔𝜔𝑗𝑗 (𝑢𝑢
�⃗𝑗𝑗 − 𝑢𝑢
�⃗𝑖𝑖 ). ∇𝑊𝑊𝑖𝑖𝑖𝑖 , (8)
𝑗𝑗∈℘(Ω)
𝑑𝑑𝑥𝑥
���⃗𝚤𝚤 (9)
= �����⃗,
𝑣𝑣0𝚤𝚤
𝑑𝑑𝑑𝑑
𝑑𝑑𝜔𝜔𝑖𝑖
����⃗∗ − ��𝑣𝑣�⃗�.
= 2𝜔𝜔𝑖𝑖 � 𝜔𝜔𝑗𝑗 �𝑣𝑣 𝚤𝚤 ∇𝑊𝑊𝑖𝑖𝑖𝑖 , (10)
𝑑𝑑𝑑𝑑
𝑗𝑗
𝑣𝑣 ∗ and 𝑃𝑃∗ represent solutions of the linearized Riemann problem. With such a
scheme, particles are moved with a speed-law 𝑣𝑣 0𝚤𝚤 (see Eq. (9)) in order to
�����⃗
disrupt particle structures (See Oger, 2015).
The dynamic particle refinement algorithm used in this article was initially
proposed by Feldman (2007) and improved by Barcarolo (2014). In this
algorithm, when a particle needs to be refined (mother particle) it is divided
into a finite number of smaller particles (called daughters particles) following a
predefined pattern.
Presented at the NAFEMS World Congress 2019 Québec City, Canada | 17-20th June 2019
© NAFEMS 2019 REPRODUCTION AND REDISTRIBUTION PROHIBITED nafems.org
Particle refinement process: mother (red) particles are split into four
daughters (blue) particles.
We also define a mass ratio 𝜆𝜆𝑑𝑑 ∈ ]0,1] which gives the mass of each daughter
particle:
𝐷𝐷
� 𝜆𝜆𝑗𝑗 = 1, (15)
𝑗𝑗
Presented at the NAFEMS World Congress 2019 Québec City, Canada | 17-20th June 2019
© NAFEMS 2019 REPRODUCTION AND REDISTRIBUTION PROHIBITED nafems.org
where 𝐷𝐷 stands for the number of daughter particles of each mother particle
(𝐷𝐷 = 4 in Figure 1 for example).
5. Numerical validation
Kleefsman (2005) proposed the experiment test case of a dam break impacting
an obstacle. Despite this test case displays three dimensional effects in a first
attempt, we treat this flow as 2D. This approximation is acceptable namely at
the first instants of the impact. We therefore performed a 2D simulation (see
Figure 2) and compared our results for various discretizations with
experimental pressure sensors.
Presented at the NAFEMS World Congress 2019 Québec City, Canada | 17-20th June 2019
© NAFEMS 2019 REPRODUCTION AND REDISTRIBUTION PROHIBITED nafems.org
Figure 5 (a) and (b) show the evolution of pressure sensor P1 in comparison
with the experimental signal. We clearly identify the gain using Particle
Refinement compared to the coarse simulation on Fig. 5 (a). A good agreement
between particle refinement and fine spatial resolution is obtained, which can
be clearly identified using a zoom (Fig. 5 (b)). The use of Particle Refinement
gives the same trend on the P3 sensor as coarse and refined simulations (Fig. 5
(c)). The differences observed between the numerical results and experience
may attributed to 3-D effects. It may be noted that the simulation performed by
Kleefsman (2005) gives the same trends, namely a result slightly higher for P1
and slightly lower for P3.
Presented at the NAFEMS World Congress 2019 Québec City, Canada | 17-20th June 2019
© NAFEMS 2019 REPRODUCTION AND REDISTRIBUTION PROHIBITED nafems.org
Presented at the NAFEMS World Congress 2019 Québec City, Canada | 17-20th June 2019
© NAFEMS 2019 REPRODUCTION AND REDISTRIBUTION PROHIBITED nafems.org
Saving CO2 emission is the key issue for current internal combustion engine
developments. Lubrication efficiency in the gear box is one of contributors to
limit CO2 emission by limiting the on-board mass of oil. Consequently,
developers of the lubrication system need to minimize the volume of oil in the
gear box without any penalties on reliability of the system.
The description of the lubricant flow in the gear box should be addressed by
any numerical technics for free surface simulation.
Bearings
Gears G
For lubrication, the wheel is used as a pump to put the oil in the collector.
Thus, the oil mass flow is split through dedicated holes, partly to the gears,
partly to the bearings.
Presented at the NAFEMS World Congress 2019 Québec City, Canada | 17-20th June 2019
© NAFEMS 2019 REPRODUCTION AND REDISTRIBUTION PROHIBITED nafems.org
7. Validation
The simulation time is about seven days on 70 CPU cores for three seconds
real time and a SPH-flow simulation needs about 1.5 million of SPH particles
for 1.5 liter of oil. The use of local particles refinement (see Figure 8) helps to
achieve higher fidelity in zone of interest such as bearings or holes in the oil
collector. Consequently, covering all the simulation plan and complying with
the time schedule of the development means that all the simulations must be
run simultaneously, so a huge CPU capability is required.
Presented at the NAFEMS World Congress 2019 Québec City, Canada | 17-20th June 2019
© NAFEMS 2019 REPRODUCTION AND REDISTRIBUTION PROHIBITED nafems.org
The second step of the validation process in on real gear box and the test is
reliability test, maximum power, maximum speed, up to break the system.
8. Conclusion
In recent years, the limitations of the original SPH method have been overcome
thanks to several innovative techniques briefly introduced in this paper. These
are implemented in the code SPH-flow made it possible to make the transition
from an academic method to a solver for complex engineering problems, here
lubrication applications.
The SPH method is suitable for dynamic flows and complex interfaces. In this
paper, the comparison of simulated oil pathway to experimental results by
transparent casing demonstrate that SPH-flow simulations can predict
accurately the main oil flows during splash lubrication highlighting the key
parameters of gearbox lubrication mechanism, contributing to reliability and
CO2 emission saving.
Presented at the NAFEMS World Congress 2019 Québec City, Canada | 17-20th June 2019
© NAFEMS 2019 REPRODUCTION AND REDISTRIBUTION PROHIBITED nafems.org
9. References
Monaghan J.J. (1994). Simulating free surface flows with SPH: Journal of
Computational Physics. 110: 399-406.
Tait P. (1888). Report on some of the physical properties of fresh water and
sea water: Physical Chemistry. 2.
Oger G., Marrone S., Le Touzé D. (2015). SPH accuracy improvement through
the combination of a quasi-Lagrangian shifting transport velocity and
consistent ALE formalisms: Journal of Computational Physics.
Feldman J., Bonet J. (2007). Dynamic refinement and boundary contact forces
in SPH with applications in fluid flow problems: Int. J. Numer. Meth. Eng.
72:295324.
Chiron L., Oger G., Le Touzé D., De Leffe M. (2018). Analysis and
improvements of Adaptive Particle Refinement (APR) through CPU time,
accuracy and robustness considerations: Journal of Computational Physics.
354:552–575.
Presented at the NAFEMS World Congress 2019 Québec City, Canada | 17-20th June 2019