Transcript Order Analysis

RotasPro Noise Analysis
Transmissions, Gearboxes and Axles
1.11.2006
1
Discom Industrial Measurement Inc.
Cabin noise
Base System
Gearboxes and Components
APAS-II
ROTAS-Mobil
MESAM 4
The company was founded in 1985. It is situated in Göttingen in the heart of Germany. Since 1989 the
main product are measurement system for the acoustical quality analysis. There are 24 employees and
more than 400 test systems deployed.
Base system are ROTAS and MESAM-4 (a joint development with DaimlerChrysler). Build on top of
them are APAS for the cabin noise analysis, ROTAS-GP for the analysis of gearboxes and transmissions,
ROTAS-ZP for the gear test, ROTAS-TMO for the test of tapered roller bearings.
1.11.2006
2
Discom: Customers and Applications
American Axle, USA
Front and Rear Axles, 5 Systems
Jatco, Japan
CVT Transmissions
Bentley, UK
Cabin Noise Analysis APAS, 2 Systems
GM, USA, Opel D and A
Gearboxes 14 Systems and Axle Testers 25
Systems
Borg-Warner, USA,UK Transmissions und
Axles 8 Systems
Renault, Peugeot
Transmissions, 5 Systems
DaimlerChrysler D, US Cooperation MESAM4,
gearboxes, transmissions and engines.
36Systems
Robert Bosch Starters, injection pumps,
production database 6 Systems
First Automobile Works, China
Gearboxes 1 System
Fiat, Iveco, Tata
Manual Transmissions 6 Systems
Ford, D
Gear Testers 4 Systems
Gearbox, SP
gearboxes, gear testers 12 Systems
1.11.2006
SKF D, USA, India, Ukraine
Tapered Roller Bearings, 26 Systems
SAIC Shanghai Automobile Internatiol Corp.
Gearboxes test systems, 4 Systems
Skoda, Cz
Gearboxes, 7 Systems
SFT, Magna Stey Powertrain
Axles and gearboxes 28 Systems
Graziano, Italy Gearboxes and Axles for Audi,
Aston Martin, Ferrari, Lamborghini, Maserrati
Tongil, Korea
Axles, Durability test Stand, 1 System
Hyundai, Korea
Gear tester 2 Systems
Volkswagen D,SA,AR, China,B,SK,BR
Gearboxes, Transmissions, Gear Testers, Cabin
Noise Production Database, 155 Systems
Linamar, Canada
Axles 2 Systems
ZF D
Gear Testers 4 Systems
3
Rotas NVH Testing
1.11.2006

End of line test with vehicle correlation and the ability to find
production errors

Separation of noise components from inner wheels of
transmissions

Calibrated evaluation for comparison of the production
process with car experiences

Simultaneous measurement and evaluation of all
transmission order components

Recognition of nicks on gears and bearings

Trainable system in combination with fixed limits

Test results stored in central data base

Intranet evaluation of test results for groups of test benches
4
ROTAS Test Stand Configuration
Parameter- und
Limit Database
Sensors
Rotas
Analysis
Program
Test Stand
Control
Result Database
Archives, SQL-DB
Presentation
1.11.2006
Web Statistics
5
Rotas Analyzers
19”-Industrial PC.
3 DSP’s, Speed Card,
Monitor output
2.6 GHz P4
BlackBox, 3 DSP’s Frontend,
Speed Card,
Monitor output
Rotas Mobile, 12 Volt Car Supply
3 DSP‘s, Monitor output, Laptop
1.11.2006
6
Analysis of transmission noise
RotasPro separates the noise into the
different sources inside the transmission like
input, intermediate and output shaft. I
addition, the the total noise is evaluated.
A main noise source is the tooth mesh of the gears.
A prominent method is the calculation of the spectral components of the
noise. The spectrum shows the characteristic of the noise in form of a
curve.
For transmissions, spectral components exist for tooth mesh and its
harmonics.
Noise
sources
The spectra are compared with limit curves. If the limit curve is
exceeded, a defect can be assumed.
In addition, temporal measurement give Crest, Peak and RMS values,
which are used for the identification of nicks on the transmission parts.
1.11.2006
7
Order Analysis
Exact order analysis is accomplished by
generating of the same amount of samples per
revolution of a part, independent of its
revolutionary speed.
Input
One rev
Digital resampling allows for the processing of 3
to 5 shafts of a transmission.
Input
From this resampled time domain signal, spectra
evaluation produces order components. These
are the amplitudes of sine components, which
are cyclic with the revolution of the part
One rev
Input
One rev
Constant number
of samples per
revolution
1.11.2006
8
Revolution Synchronous Transmission Analysis
The transmission noise is the sum of
the noise originating from the
individual mechanical components.
Input Shaft
Interm. Shaft
For the gear noise components, the
individual sources can be isolated by
the transmission ratio.
Output Shaft
Synchronous
order analysis:
The signals are
acquired synchronous
to the inner shafts.
Acoustical
Stroboscope
Input Shaft
1.11.2006
Interm. Shaft
Output Shaft
9
Synchronous Averaging I
Signal
For every shaft, averaging is performed
synchronous to its revolution.
Signal components of the synchronous shaft are
enhanced, other noise components are
attenuated.
One shaft is shown. Rotas can do the processing
for 3 to 5 shafts simultaneously.
Background
+
+
Signal
+
Background
+
Usually, only one sensor is needed.
Signal
The synchronization can be done for inner
shafts, for which only the transmission ration is
known.
+
Background
Summe * 1/n =
Average:
Signal
enhanced
1.11.2006
10
Synchronous Averaging II
g
Nicks are shown in the synchronous
channel that is synched with the offending
shaft.
The nick detection is based on the Crest
value
Zw An
16.0
8.0
0.0
-8.0
-16.0
g
Crest = Peak value/RMS value
Zw Ab
16.0
8.0
0.0
-8.0
-16.0
g
TsAb
16.0
8.0
0.0
AG4 V4.0 Typ 39 Unit 67 04.Jun'2000,17:10:16
------------------------------------------------------------------Code Description
Gear Is
Lim
Av Unit Kx
448 NICK 3.-Gear / Up
3-Z 16.8/ 12.0/ 9.5 Crest ZwAn
--------------------------------------------------------------------
1.11.2006
-8.0
-16.0
400 800 1200 1600 2000 2400 2800 3200 3600m Rev
11
Separation of Gear Errors
The acoustical signal holds the
components of both gears.
Knowing the transmission ratio, the
periodicity of each gear can be found.
Gear mesh components only depend
on the pairing of the gears.
Eccentricities and surface errors can
be separated because they have a
cycle that corresponds to the
originating gear. The following errors
can contributed individually:
•
•
•
•
•
Eccentricities,
Deviation from circular shape
Tooth spacing
Surface waves (Ghost Orders)
Nicks
Z1
16
Z2
20
20
Z1
The gear mesh component
depends on the pairing
16
16
Z2
20
20
1.11.2006
16
Eccentricities and surface
errors can be contributed
individually.
12
Time Domain Versus Order Domain I
Gear with 5 teeth, 4 revolutions
Sine
smooth gear contact
Triang
harsher gear mesh
Sawtooth
harsh gear mesh
V
V
2.0
Oran.SK1
2.0
Oran.SK1
V
2.0
Oran.SK1
1.0
1.0
1.0
0.0
0.0
0.0
-1.0
-1.0
-1.0
-2.0
-2.0
-2.0
400 800 1200 1600 2000 2400 2800 3200 3600m Rev
400 800 1200 1600 2000 2400 2800 3200 3600m Rev
400 800 1200 1600 2000 2400 2800 3200 3600m Rev
Fourier transform: Order spectrum
dBV
100
Oran.SK1
5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0
Order at number of teeht: 5
Base order H1
1.11.2006
dBV
100
Oran.SK1
dBV
100
Oran.SK1
90
90
90
80
80
80
70
70
70
60
60
60
Ord
5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0
Ord
Harmonics at odd multiples
of nr of teeth (5)
H1, H3, H5, ..
5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0
Ord
Strong harmonics of number of
teeth (5)
H1, H2, H3, ..
13
Time Domain Versus Order Domain II, Modulations
Gear with 5 teeth, 4 revolutions
Sine with modulating sine
Smooth gear mesh with
eccentricity
Triang with modulating sine
harsher gear mesh with
eccentricity
Triang with modulating triang
harsher gear mesh with
several eccentricities
2.0
Oran.SK1
V
V
V
2.0
Oran.SK1
2.0
Oran.SK1
1.0
1.0
1.0
0.0
0.0
0.0
-1.0
-1.0
-1.0
-2.0
-2.0
-2.0
400 800 1200 1600 2000 2400 2800 3200 3600m Rev
400 800 1200 1600 2000 2400 2800 3200 3600m Rev
400 800 1200 1600 2000 2400 2800 3200 3600m Rev
Fouriertransformation: Ordnungsspektrum
dBV
dBV
100
Oran.SK1
5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0
Oran.SK1
dBV
100
Oran.SK1
90
90
90
80
80
80
70
70
70
60
60
60
Ord
Order at nr of teeth H1 with side
bands +/- 1 order
1.11.2006
100
5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0
Harmonics H1, .. with
side bands +/- 1 order
Ord
5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0
Ord
Harmonics H1, .. with
side bands +/- n.Order
14
Synchronous Averaging III
Revolution synchronous averaging gives
periodic (cyclic) signals. This corresponds
to the cyclic nature of the gear sets.
dBV
GM
These signals can be transformed into the
spectral domain without any time domain
leakage windows.
Mix
This allows for high spectral resolution.
Eccentricities can be easily distinguished
from the gear mesh orders. The noise
components can be attributed to their
origins.
Antrieb
2 *GM
80
Ecc..
Ghost
Orders
70
60
50
Blue: Conventional order spectrum with
Kaiser Bessel Window
40
Green: Revolution synchronous order
spectrum without window function
50 60 70 80 90 100 110 120 130 Ord
1.11.2006
15
Order spectra of a lay shaft transmission
dBg
Limits:
red
Input H1
AW
Input H2
110
100
SK1-lim
90
O-Spectra:
black
Input
80
70
dBg
Input: 26 Z
VZ1: 35 Z
VZ 4: 43 Z
5. Gear 27 Z
i = 1.1764
VZ1 H1
VGW
VZ5 H1
VZ1 H2
VZ5 H2
110
100
SK2-lim
90
80
VGW
70
dBg
5. Gg H1
HW
OW 5. Gg
5. Gg H2
110
100
SK3-lim
90
80
HW
70
dBg
-12 dB
Mix
-10 dB
110
100
Mix-lim
90
80
Mix
70
8.0
1.11.2006
16.0
24.0
32.0
40.0
48.0
56.0
64.0
72.0
Ord
16
Evaluation of Order Spectra
The order spectra of the synchronous channels
and of the mix channel are compared with a
limit curve. Every order of the limit curve
has an underlying error code. This error code
is generated from transmission ratios in a
semiautomatic fashion. If the limit curve
is exceeded, the error code gives rise to an
error message in plain text.
dBg
VGW
105
VGW-lim
The limit curves consist of portions, which
are generated by a learning process and of
portions which are set to fixed values.
90
75
60
45
10
20
30
40
50
60
70
80
90
Ord
The learn rnode is used for those spectral
portions, where there are results from driving
experiments available yet. Generally this
holds true for tooth spacing problems,
ghost orders and bearing noise (in the
mix channel).
Fixed limits are adopted from driving experiments,
usually for the gear mesh and for the side bands.
1.11.2006
17
Measurement Protocol
A protocol is generated for every unit.
It holds information about unit type,
serial number, time of measurement
and defect report.
There are three kinds of report:
Short:: serial numbers only for good
units, defect report for failing units.
Normal: Report line for every unit
Long: All measurement values in
tabular form.
Around 300 measurement reesults
are stored in a protocol database for
statistical evaluation. 5000 units use
about 20 Mbyte of disk space.
The test stand control system can
request several protocol variants
for own recording purposes.
1.11.2006
R2528 Typ R25 Seriennr. 1480889 29.May'2000,06:12:10 Getriebe n.i.O. ***
-------------------------------------------------------------------------------Code Fehlertext
Gang
Ist Grenz Mw
E
O/Tp Kx
6 Rücklaufrad laut
R-Z 98.5/ 93.0/ 88.4 dBg
29.00 RLR
13 Rundlauffehler VZ I
2-Z 102.3/ 97.0/ 91.9 dBg
72.00 VGW
-------------------------------------------------------------------------------R2528 Typ R25 Seriennr. 1480984 29.May'2000,06:27:05 Getriebe n.i.O. ***
-------------------------------------------------------------------------------Code Fehlertext
Gang
Ist Grenz Mw
E
O/Tp Kx
3 VZ I laut
1-S 108.2/106.0/101.2 dBg
70.00 VGW
2 Antriebswellenrad laut
1-S 109.0/106.0/101.2 dBg
52.00 AW
-------------------------------------------------------------------------------R2528 Typ R25 Seriennr. 1480992 29.May'2000,06:37:35 + Getriebe i.O. +
-------------------------------------------------------------------------------R2528 Typ R25 Seriennr. 1480986 29.May'2000,06:56:04 Getriebe n.i.O. ***
-------------------------------------------------------------------------------Code Fehlertext
Gang
Ist Grenz Mw
E
O/Tp Kx
12 Rundlauffehler Rad AW
1-S 105.2/101.0/ 91.6 dBg
55.00 AW
-------------------------------------------------------------------------------R2528 Typ R25 Seriennr. 1480985 29.May'2000,07:07:58 + Getriebe i.O. +
-------------------------------------------------------------------------------R2528 Typ R25 Seriennr. 1480990 29.May'2000,07:18:56 + Getriebe i.O. +
-------------------------------------------------------------------------------R2528 Typ R25 Seriennr. 1480987 29.May'2000,07:43:06 + Getriebe i.O. +
-------------------------------------------------------------------------------R2528 Typ R25 Seriennr. 1480988 29.May'2000,07:53:45 + Getriebe i.O. +
-------------------------------------------------------------------------------R2528 Typ R25 Seriennr. 1480991 29.May'2000,08:04:02 + Getriebe i.O. +
-------------------------------------------------------------------------------R2528 Typ R25 Seriennr. 1480991 29.May'2000,08:05:00 + Getriebe i.O. +
------------------------------------------------------------------------------R2528 Typ R25 Seriennr. 1480650 29.May'2000,08:26:50 + Getriebe i.O. +
-------------------------------------------------------------------------------R2528 Typ R25 Seriennr. 1481405 29.May'2000,11:17:24 + Getriebe i.O. +
--------------------------------------------------------------------------------
18
Spectral Limits
Hats for gear mesh orders and their side bands
dBg
100
MaxBnd
80
MinBnd
60
StdDev
40
AW-avg
20
SK1-lim
10
20
30
40
50
60
70
80
90
Ord
Limit curve from Average + Offset + n times standard deviation. Overridden by minimum
and maximum polygons and by the fixed hats.
1.11.2006
19
Learning
What is learned
For all measurements, the following statistical
properties are computed:
Average (AV) und
Standard deviation (STD)
From these numbers, a limit L is generated
according to
Learning methods
A new learn consist of a base learn (usually 5
units), where all units are accepted that are
below maximum polygons and below hats.
The base learn is followed by an additional
learning (usually 100 units). Here every unit is
evaluated against the already established limit
curves. Only error free units will be added to
the learning sample.
L = AV + Offset + n* STD.
The offset and the multiplier of the standard
deviation is selectable. For spectral measures,
an offset of 5 dB and a std-factor of 3 is
normally used.
Maximum and minimum values set upper and
lower bounds for the limit L.
1.11.2006
Another available choice consists of a indefinite
learn process with a selectable time constant.
During the whole learning process, the fixed
limits apply. This assures, that all known data
(like limits from car experiments) will be used
as a guidance during the process.
20
Cabin noise components
Noise sources
There are many different
transfer paths for the sound
and the vibration from the
gearset to the drivers ear.
Each path has its own
frequency dependent
transfer function
Transfer paths
Noise from the motor, wind
and tires may mask certain
frequency bands from the
gears.
These effects induce the
speed, torque and vehicle
dependent noise perception
of the gear sets.
1.11.2006
21
Averaging over microphone postions
Speed
M1
M2
M3
To compensate for the cabin resonance's, four
microphones are positioned within the cabin:
Front left, middle and right position, one in the
rear. The energy of the microphone signals is
averaged in the spectral order domain.
M4
Accel
Order
Resamp
One sensor is mounted on the transmission to
allow for the simultaneous analysis of the gearbox
vibration data.
FFT
Average
+
Spectrogramm
Ordertracks
1.11.2006
22
Order spectrograms
Order spectrograms are recorded for the gearbox
vibration signal and for the cabin sound field.
2) Order spectrum at 2241 RPM
From these order tracks for arbitrary orders and order
spectra for arbitrary speeds can be generated:
Order track for the 65. order
1.11.2006
1) Ordnungs-Spektrogramm 4. Gang Zug
23
Vibration and Cabin Noise, Spectral Cuts
12 dB
Spectral Cut Vibration: 50th order of 100 dBg at 2700 RPM
1.11.2006
Spectral Cut Cabin Noise: 12 db higher than background
at 2640 RPM/110 Km/h
24
Vibration and Cabin Noise, Order Cuts
1.11.2006
25
Order Tracking





The critical noise sources in a car are
usually the gear mesh orders and their
sidebands.
These noise sources are highly speed
dependent. Detailed speed dependent
evaluation uses the track of mesh orders
and sidebands over speed.
The gear mesh H1 base order and its
harmonics are tracked over speed and
compared with speed dependet limit
curves.
The track form depends on torque,
drive/coast and on the resonances within
the gearbox and the test stand.
Within speed bands, average and
maximum amplitudes can be evaluated
and compared to limits. The single values
are ideal for statistical analysis over time.
Speed band
Tracked Value
H1 Max Bd1
H1 Aver Bd1
1.11.2006
Curve 1
79.1 dB
76.2 dB
Curve 2
75.5 dB
71.9 dB
26
Shift force evaluation



Shown are measurements with the GSW
shift roboter, which measures forces and
displacement directly at the transmission
shift mechanism.
B
Shifting effort is shown versus
displacement
A
C
B
C
Positive forces: Shift into a gear
B

Negative forces: Shifting out of gear

A: Synchronisation work when shifting into
a gear. Good repeatability

B: Shifting through after synchronisation.
Varying force.

C: Shifting out. Small forces for good
transmissions
D: Synchron ring missing

A
1.11.2006
A
C
B
D
B
C
D
A
C
A
A
B
C
27
Evaluation of Shifting Effort

Evaluation of shifting effort versus lever
displacement.

The synchronization is evaluated via the
work that is needed (F*s). Therefore, the
area of the curve above a window A is
computed (work = force * displacement) .
D: a transmission without a synchron ring.

The maximum force during shifting into a
gear are limited with the polygon B

The maximum forces when shifting out are
limited with a polygon C.

End position: a settable end position must
be reached when shifting is finished.
B
A
D
C
Schaltkraft über Weg für 150 Seriengetriebe. Bewertungsfenster
erkennen fehlende Synchronringe und Schwergängigkeiten.
1.11.2006
28
Management of Multiple Test Benches
Test Benches
Data
Parameter
Parameter Setting and
Statistical Evaluation
PC with
Internet
Explorer
Server
Intranet
Central archive of test results
1.11.2006
Parameter setting and statistical evaluation
from every PC with intranet access.
29
Rotas Data Flow
Rotas
Analyzer
Work Group PC has presentation software installed
Can use Intranet statistcis via Internet Explorer
Test Stand
Measurement
Archives
Rotas
Presentation
Measurement
Database
Intranet Statistics
Server or Test Stand
1.11.2006
Work Group PC
30
Measurement Transfer over Network
Parameter
Database
Local Copy
Measurement
Database
SQL Server
2000
ROTAS
Analyzer
Measurement System
1.11.2006
Archive Files:
One Test Run
Archive Files:
One Test Run
\inbox:
Archive Files:
One Test Run
\outbox:
Collector Service
Server
31
Order Spectra




Spectra are stored for the individual rotors
in the transmission
An additonial Mix spectrum represents the
order amplitudes of all parts combined
The spectra are either peak hold
amplitudes or averaged amplitudes over a
ramp
Limits can be self learned or manually set.
The amplitude of gear mesh orders can
be set individually by a database
1.11.2006
32
Spectral Statistics
Order spectra of production units can be displayed in
Campbell or in 3-D graphs. Cuts in the order or in the
unit direction show order amplitude versus units or
the spectrum of one unit. The pictures show the order
spectra of 300 production units. A ghost order of 118
is present in the first 40 units tested.
1.11.2006
33
Order Tracks
During the ramps, spectral
components are evaluated over
the speed.
Usually, the main gear mesh
orders and their harmonics are
tracked.
In addition, energy in order bands
may be tracked as well.
Track limits (shown in red) can be
self learned or set manually.
1.11.2006
34
Complete Data Report of a single Unit
Sensor Group 1
Macro controlled
display of archived
data.
The database is
used for searching of
serial numbers, test
results etc.
Then the
presentation
software displays a
full measurement
report
1.11.2006
35
Intranet Statistics
The information of
the SQL
measurement
database can be
retrieved via the
intranet.
The production
results and the
measurements can
be seen online.
An arbitrary time
interval can be
selected for the
analysis.
1.11.2006
36
Statistical Evaulation for Types and Benches
Test results The
evaluation can be
done over gearbox
types or over test
benches.
Any time interval can
be specified.
1.11.2006
37
Measurement Statistics and Bench Comparison
Over 300
individual
measurement
results can be
evaluated.
The
comparison of
test benches
makes it easy
to administer
large
production
facilities.
1.11.2006
38
Reject Analysis, Top n and Type Specific
The results are
shown both in tabular
and in graphical form
1.11.2006
39