Priyatosh Barman, Six Sigma black belt at Caterpillar explains: "Some
dealers had informed us that certain hydraulic pumps on the larger
vehicles were failing before their expected shelf life. Usually, the
larger the machine, the more analytical sophistication is required to
ensure reliability so we turned to CFD (computational fluid dynamics) to
try to isolate the problem and prevent its recurring."
When the engineers investigated, they found that there was too much air
in the flow through the hydraulic pump, causing it to fail more
frequently and preventing the vehicle from functioning to maximum
capacity. The over-aerated flow was coming in from another part of the
hydraulic system, a partially filled tank which receives the return flow
of the fluid from the implement pump. The flow arrives into the tank at
high speed and had previously been slowed using perforated rectangular
baffles. It was found that although these baffles were to a certain
extent calming the flow and weakening its circulation, a high level of
aeration was still occurring, leading to the pumpís failure.
The over-aerated flow was being carried, via suction lines, to both the
implement and fan pumps, increasing the chance of cavitation in both
pumps. Once this had been identified as the source of the problem
STAR-CD, CFD simulation software from The CD adapco Group was used to
find ways to calm the strong fluid circulation inside the tank; reducing
aeration and further problems.
Hydraulic simulation is not traditionally assisted by CFD, dealing as it
does with fluid pressure rather than flow, but the Caterpillar design
engineers could see that in this case, CFD might offer them a more
detailed look at the problem, as well as help to design the solution.
"We started by using STAR-CD to look at exactly what the fluid was doing
inside the tanks. The software allowed us to predict the fluid
circulation within the tank and to find the fluid pressure and speed
distribution for standard flow rates," continues Mr Barman. Using this
information to create a CFD model, Caterpillar was able to examine all
the conditions within the tank, including the details of the perforated
baffles, and identify how each affected the flow around the system.
Mr Barman continues: "As well as modelling the way in which the
established solution was working, we were also able to model three
alternative designs in order to find the optimum solution to this
problem." The options included a variety of baffle designs. Perforations
were added or removed on the CFD model and the results recorded.
"STAR-CD was able to predict significant differences of flow circulation
and other flow parameters, in each case giving us an accurate picture of
the most appropriate design," says Mr Barman. "The various models
allowed us to look at what wasnít working properly and how we could
improve the design of the tank and baffles in order to avoid any further
problems with the hydraulic system."
The size of the flow circulation was found to have an impact on the
level of aeration in the tank and comparisons made using the STAR-CD
software demonstrated the optimum elements of each design. The software
led the designers to a solution that would not necessarily have been
found without its help. To solve the problem, the baffles were
surrounded with a porous circular tube. This design was found to
distribute the fluid jets from the baffle holes in a more uniform
manner, decreasing the flow circulation within the hydraulic tank.
A further improvement was discovered during the simulation process: as
well as making the flow more uniform, the porosity of the baffles was
found to play a major role in controlling the pressure within the tank.
"Using the STAR-CD software gave us three major benefits which were
invaluable in finding the best solution to this problem," concludes Mr
Barman. "The software allowed us to find and address the problem
quickly, to spend very little time on physical prototypes and, in the
end, less money was needed to resolve the problem and improve the
overall design of the tank."
Velocity contours of the baseline model.
Velocity contours of prefered model.