A sports car is supposed to sound like one, purring at idle and growling
under heavy acceleration. To owners of high-performance cars, this constitutes
music. Occasionally, however, a false note is struck. When that happens before
a car goes into production, engineers get to work, and quickly.
A false note was detected in the new Porsche Boxster, a handsomely styled
sports car designed like its ancestors for power, grace, and smooth handling.
Porsche engineers in the research and development center in Weissach, near
Stuttgart, Germany, discovered a harsh and annoying rattle under the Boxster's
hood at around 2,250 rpm.
The rattling sounded as if a metal
washer had been dropped into the manifold. Weissach engineers soon traced the
noise, around 730 Hz, to the plenum of the air intake manifold on top of the
engine. At that time, the part was undergoing final pre-production checkout and
there were just a few days to solve the problem. One of the companies
responsible for development of the product concept and its engineering as well
as actual production, Mark IV Automotive Systémes Moteurs, requested the
help of its partner, DuPont. DuPont engineers met the challenge with high-speed
detective work.
The manifold is made with a DuPont glass
fiber-reinforced thermoplastic called Zytel® 70G35 HSL Nylon. As suppliers
to Porsche, Mark IV and DuPont accepted considerable engineering
responsibility. DuPont and Mark IV shared responsibility for strength
calculations around the parts' flanges, mold flow predictions, and interior
pressure simulations. Intake manifolds must withstand the force of engine
backfires.
By standards of volume in the automotive industry, the
Boxster implementation is relatively small, only about 17,000 cars a year with
two manifolds on each car. Popular car models are built at rates six to twelve
times higher.
"But this was a key win for DuPont and Zytel," said
Glenn Philip Sievewright, DuPont's lead engineer on the Porsche Boxster
project. "This application shatters forever the myth that automakers must give
up sound quality to gain the many advantages of a thermoplastic air-intake
manifold." Sievewright works in DuPont's automotive R&D facility in Hemel
Hempstead, England.
Porsche went back to DuPont and Mark IV with a
noise-vibration-harshness (NVH) problem. Because time was short, the German
engineers came well armed with data. Tests at Porsche pinpointed the source as
air pulsations in the manifold's plenum chamber. This in turn was traced to an
increase in the inlet and exhaust valve overlap, a function of the engine's
camshaft timing. Porsche had also eliminated engine vibration in the engine
itself as a cause. There were no natural frequencies in the problem area around
730 Hz in the Boxster's 2.5 liter, a flat-head, 6-cylinder engine.
With the intake manifold on the verge of going into production, Mark IV and
DuPont engineers wanted every possible assurance that they were indeed on the
right track. This required a multiple analysis approach combined with
verification by modal finite element analysis (FEA). The software used was
ANSYS/Mechanical™ design verification and optimization software from
ANSYS, Inc., Canonsburg, PA, USA, plus acoustic holography and laser scanning
analysis packages.
"We couldn't just do anything we wanted,"
Sievewright recalled. "Our range of options was limited by the fact that
Porsche engineers had already signed off on all the parts designs, all the
under-hood placements, the spacing between components, air flows, cooling
flows, and so on.
"Though we had only seven working days to solve the
problem, the engineers took the time to do correlations between Porsche's tests
and our own," he noted. "Only then did we launch a detailed analysis of the
manifold to find out exactly what was happening."
Acoustic holography
and laser scanning analysis were used in combination, a well-tested technique
at DuPont for identifying sources of unwanted noise and ways to eliminate it.
ANSYS was used to correlate the results from acoustic holography and laser
scanning analysis. But those techniques are essentially diagnostic instruments,
Sievewright pointed out. Modal FEA is needed if their data is to undergo any
kind of intensive examination. ANSYS has a long history of success at DuPont.
"We have done hundreds of analyses with it, including the original analyses
when the Boxster manifold was developed," Sievewright noted.
But those
tests were primarily static and dynamic structural. For the NVH analyses, the
geometry of the original manifold design was read in as an IGES file. The
surfaces were extracted and meshed, and DuPont went at it from there. For
hardware, engineers used a SPARCstation from Sun Microsystems Inc., a UNIX
machine.
Acoustic holography was used to reveal which areas of the
manifold were resonating. The manifold was set up on a shake test rig and
vibrated at 730 Hz. Sound pressure levels at 15 points on the plenum's upper
surface were scanned and plotted. "The movement was as much as half a
millimeter," Sievewright said, or roughly 0.020 inch. "You could actually see
it." While still on the shaker rig, the part was laser scanned, too. The
extreme accuracy of laser scanning quickly established which sections of the
plenum surface vibrated most. The laser scans generated a plot of the mode of
the resonant frequencies," he added. "That gave us a modal picture of the
vibration. Acoustic holography just gives us vibrations at specific
points."
Modal analysis with ANSYS was brought to bear on the problem.
DuPont used a constant G-force and analyzed the part at various frequencies
across the range to further pinpoint the fix. "The surface velocities from the
modal analyses compared very well with those from the laser scanning and were
verified with ANSYS," said Sievewright.
The DuPont team found that
noise levels were highest at the area of the plenum chamber with the highest
surface velocities. "This indicated that the surface vibration energy of the
parts at resonance was emitted as that rattling noise," Sievewright said.
DuPont also found that noise emitted from the part was directly proportional to
the surface velocity.
The tests were run in parallel rather than
sequentially. "Any of three analyses probably could have told us what we needed
to know," Sievewright noted. "We did all three to make sure we got it right the
first time." Some acoustic analyses were performed with a specialized package
called Comet/Acoustics from Automated Analysis Corp. (AAC), Ann Arbor, MI. AAC
is an ANSYS Enhanced Solutions Partner, as well as an ANSYS Channel
Partner.
It turned out that the noise came from that part of the
intake manifold right under the letters SCH in PORSCHE. Fixing the problem was
straightforward: a quick redesign of the manifold's plenum to thicken the
offending surface.
ANSYS was also used to determine how difficult the
changes would be to make. "We could not change the curvature of a surface
because of the time that would be needed to modify the tooling," Sievewright
said. "But adding about one and a half millimeters of material to thicken the
plenum surface, was very effective." The mold maker simply machined away some
metal from the surface of the tooling.
This led to a 10-decibel (dB)
reduction at 730 Hz as well as a 10-dB reduction at a resonating frequency of
890 Hz. "We also got a smoother slope to the frequency spectrum between 600 and
925 Hz resulting in less harshness and lower overall sound levels." This
correlated well with the noise curve from the moisture conditioned part and
Porsche engineers signed off on the solution.
DuPont and Mark IV
quickly produced some prototypes and shipped them to Porsche for final testing.
Sievewright said that all the testing and analyses - FEA, acoustic holography,
and laser scanning analysis - plus the proposals for change and delivery of the
new prototypes were done in just seven days. "Using ANSYS let us be 99 percent
sure that we had the right solution," Sievewright said.