On noise generation and dynamic transmission error of gears

Mats Henriksson

Stockholm 2009

Royal Institute of Technology School of Engineering Sciences Department of Aeronautical and Vehicle Engineering The Marcus Wallenberg Laboratory for Sound and Vibration Research

TRITA-AVE 2009:89 ISSN 1651-7660 ISBN 978-91-7415-537-2 c Mats Henriksson, November 2009 Printed in Sweden, Universitetsservice US-AB

Preface
The work in this thesis has been carried out at the department of Aeronautical and Vehicle Engineering at the Royal institute of Technology (KTH). The funding from Scania CV AB and the Swedish Agency for Innovation systems - VINNOVA is gratefully acknowledged. I would also like to thank my supervisor Leping Feng at KTH for help and guidance in my research and my present manager Anders Abs er and my former managers Mikael P arssinen and Gunnar Strandell for allowing me to perform this work. I am grateful for all the help from my colleagues at both Scania and KTH. I would also like to thank my wife Monica and our two children, Alva and Gustav, as well as the rest of my family for their support during this work. Mats Henriksson Stockholm, November 2009

Abstract
Noise from heavy trucks is an important environmental issue. Several sources contribute to the total noise level of a vehicle, such as the engine, gearbox, tires, etc. The tonal noise from the gearbox can be very disturbing for the driver, even if the noise level from the gearbox is lower than the total noise level. The human ear has a remarkable way of detecting pure tones of which the noise from loaded gears consists of. To be allowed to sell a heavy truck within the European Union, the so called pass-by noise test must be completed successfully. The maximum noise level permitted is 80dB(A) and under certain conditions, the gearbox can be an important contributor to the total noise level. Gear noise is therefore an important issue for the automotive industry. In this thesis gear noise and dynamic transmission error is investigated. Traditionally, transmission error (TE) is considered to be the main excitation mechanism of gear noise. The denition of TE is the dierence between the actual position of the output gear and the position it would occupy if the gear drive were perfect. Measurements of dynamic transmission error (DTE) and noise have been performed on a gearbox. The measurement object was a commercial truck gearbox powered by an electrical motor. The torque used was in the normal operating range of the gearbox and the correlation between gear noise and DTE, when the torque is changed, is investigated. The result diers for dierent gear pairs and for the rst gear stage, located close to the housing, the correlation is high for most speeds. The measured DTE and noise show a poor correlation with calculated transmission error. A minimisation of TE therefore does not necessarily mean a minimisation of gear noise. A transfer function can be employed to calculate the relationship between DTE and noise. The general trend of the gear noise is an increase of 6dB per doubling of the rotational speed together with uctuations around the mean due to resonances of the system. The magnitude of the transfer function can be estimated using the amplitudes of the gear mesh orders and harmonics. Two gear pairs with similar macro geometry but dierent prole modications are investigated. Although the gear pairs have similar transmission error, the noise level display a signicantly dierent trend, further strengthening the position that transmission error is not the single most important gear noise excitation mechanism. Further analysis concludes that shuttling forces and friction forces can be more important than what is often suggested. A dynamic model including transmission error and shuttling forces is used to investigate the two gear pairs. The bearing forces show that for some frequency regions shuttling forces can be of the same order of magnitude as the forces caused by transmission error. This work highlights the importance of considering other excitations of gear noise besides transmission error when designing quiet gears. The inuence of transmission error can not be determined by investigating the gears only. A deeper knowledge of the gear system is needed in order to minimise gear noise for a specic gear design.

Doctoral Thesis
This doctoral thesis consists of this summary and ve appended papers listed below and referred to as Paper A to Paper E. Paper A M. Henriksson: Analysis of gear noise and dynamic transmission error measurements. Presented at ASME IMECE04, paper 61077 Paper B rssinen: On model of gearbox noise and dynamic L. Feng, M. Henriksson, M. Pa transmission error. Presented at ICSV12, paper 485 Paper C M. Henriksson, L. Feng: Analysis of gear noise and dynamic transmission error using a recursive Kalman lter algorithm. Submitted to International Journal of Acoustics and Vibration Paper D M. Henriksson, Y. Pang, L. Feng: Transmission error as gear noise excitation. Submitted to International Journal of Acoustics and Vibration Paper E M. Henriksson, L. Feng: A dynamic gear model including transmission error and shuttling forces. Submitted to Journal of Mechanical Design

Contribution from the author of this thesis

Paper A Performed the measurement, wrote the paper. Paper B Performed measurement, Performed analysis together with L. Feng Paper C Performed analysis, wrote the paper Paper D Performed analysis, wrote the paper Paper E Developed model, wrote the paper

Material from this thesis has also been presented at the following two conferences: Henriksson, M. and Pang, Y., 2009, Transmission Error As Gear Noise Excitation IDETC/CIE 2009, paper DETC2009-86695. Henriksson, M. and M. P arssinen,Comparison Of Gear Noise And Dynamic Transmission Error Measurements, ICSV10,2003, paper 61077

Contents
1 Introduction 1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Gear noise 2.1 Transmission error . . . . . . . . . . . . . . . . . . . . . . 2.2 Friction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Shuttling forces . . . . . . . . . . . . . . . . . . . . . . . . 3 Dynamic gear modelling 4 Measurement of gear noise and DTE 5 Summary 5.1 Paper 5.2 Paper 5.3 Paper 5.4 Paper 5.5 Paper of A B C D E main results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1 1 2 3 4 4 5 6 8 8 10 11 12 12 13 15

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

6 Conclusions and future work References

1
1.1

Introduction
Background

Noise from heavy trucks is an important environmental issue. Several sources contribute to the total noise level, such as the engine, gearbox, tires, etc. In this work, gearbox noise is investigated. The tonal noise from the gearbox can be very disturbing for the driver, even if the noise level from the gearbox is lower than the total noise level. The human ear has a remarkable way of detecting pure tones of which the noise from loaded gears consists of [1]. To be allowed to sell a heavy truck within the European Union, the so called pass-by noise test must be completed successfully. The maximum noise level permitted is 80 dB(A) [2, 3] and under certain conditions, the gearbox can be an important contributor to the total noise level. Gear noise is therefore an important issue for the automotive industry. The decreasing time for product development also creates a need for modelling and optimising the dynamic and acoustic properties of gear systems. The ability to accurately predict the dynamic and acoustic behaviour before a prototype is built can save both time and money. Also, understanding how a system behaves dynamically is crucial when trying to nd a solution to noise and vibration problems. Transmission error (TE) is considered to be the main excitation mechanism of gear noise. The denition of TE is the dierence between the actual position of the output gear and the position it would occupy if the gear drive were perfect [4]. In Dudleys Gear Handbook [5], prof. D. Houser state that Transmission error is the single most important factor in the generation of gear noise., other authors also asserts the importance of transmission error as the dominant gear noise excitation [6, 7, 8]. Additional sources beside TE can also contribute to gear noise. Research show that eects such as friction and shuttling forces can be important when analysing the dynamic behaviour of gears [9, 10, 11]. This will be discussed more in the next section. This thesis analyses the properties of the dynamic transmission error and the possibilities of using it to predict gear noise levels and the importance of other gear noise sources besides transmission error.

1.2

Motivation

A review of the available literature revealed a need for experimental study of dynamic transmission error and noise, especially for two stage gearboxes. For example, Davies et al. [12] writes in a review on gear dynamic behavior that With the increasing power of predictive methods there is a substantial lack of good correlative experimental data. Measurement of dynamic transmission error and noise was therefore the rst part of this work. The result shows large deviations from the predicted noise behaviour of the gear system based on a minimisation of transmission error. Understanding the reason for the deviation therefore became a natural continuation of the thesis.

Figure 1: Truck gearbox similar to the one used in measurements

Gear noise

Gear noise is normally divided into two parts, gear whine and gear rattle [13]. Gear whine is described as gear mesh frequency and their harmonics and is the main focus of this thesis. Lightly loaded gears have been shown to exhibit non-linear behaviour when the surfaces of the gear teeth loose contact and later collide. Rattle can be excited by external speed uctuations, for example in the timing gears of internal combustion engines where the irregular speed of the crank shaft can excite rattling of the gears. The internal excitation of the gears can also be a source of rattling and is often modelled as a transmission error excitation [14, 15]. Gear whine on the other hand originates from loaded gears. Gear whine can also display non-linear behaviour due to for example partial contact loss, which is a result of the increased mesh stiness for increased loads. Gear noise is often described as a source-path-receiver problem where the excitation occurs in the gear mesh and is then transmitted via the gear body, shafts and bearings to the gearbox housing where it is radiated as noise. In gure. 1, a two stage truck gearbox, similar to the gearbox used in the thesis, can be seen. Transmission error, friction and shuttling forces will be discussed more in the following sections, although other excitations also exist. Air and lubrication entrapment 2

occurs for high speed gears when air or lubricants get trapped between the gear surfaces and is squeezed out. Windage noise can also have an inuence for high speed gears. These excitations are only important for high speed gears and are not considered to have an impact in the tested and modelled gears of this thesis.

2.1

Transmission error

Theoretically, for two gears with perfect involutes and an innite stiness, the rotation of the output gear would be a function of the input rotation and the gear ratio. A constant rotation of the input shaft would therefore result in a constant rotation of the output shaft. Due to both intended shape modications and unintended modications, such as manufacturing errors, gears will normally not have a perfect involute shape. Also, due to a nite gear mesh stiness, there will be a motion error of the output gear relative to the input gear. This error in motion is the transmission error and is calculated according to: T E (t) = Rb,p p (t) + Rb,g g (t) (1)

Where Rb,p and Rb,g are the base radius and p (t) and g (t) are the rotation of pinion(p) and gear(g). In eq.1, transmission error is described as a length along the line of action. The line of action can be seen in g. 2. It can also be described as an angular displacement of the pinion or gear and the TE is then divided by the base radius of the pinion or gear. Measuring TE in a single ank measurement machine is often employed by gear manufacturers to verify the shape of the gear. The measurement gives no information of the gear strength, only the shape deviations. The gears are mounted with adequate spacing, representing the working conditions of the gear pair with only one ank of the gears are in contact. The transmission error and mesh stiness variation is often considered to be the primary excitation of gear noise and a minimisation of the transmission error is believed to minimise noise. Several comparisons show the correlation between TE and noise, for one speed and torque the correlation can be high, although changing the speed or torque can reduce the correlation, making it dicult to see a relationship between TE and noise. Akerblom [16] compares TE and measured noise for several gear pairs with dierent nishing methods and prole modications. For one torque there seem to be a correlation between noise and TE but for another torque the correlation is lacking. Houser et. al. also compared measured noise levels to calculated TE levels and a correlation was found in some cases but not in others [17]. N. Yildiri et. al. [18] successfully reduces noise from a helicopter transmission by changing the prole modications and thereby reducing TE for high contact ratio spur gears. Akerblom [16] presents an improved gear design with decreased transmission error which also demonstrate a decreased noise level, although the width of the new gear pair is increased and the module decreased thus increasing the total contact ratio with 36%. An increase in contact ratio has for a long time been known as an important method of reducing gear noise [19]. 3

55

50

Pitch point
45

40

Base circel

Pitch circle

35

Line of action
30

20 10 0 10 and pitch 20 Figure 30 2: Schematic view of a gear pair showing the base circles and 30 the line of action

25

2.2

Friction

In the gear contact, the sliding friction produces friction forces that acts perpendicular to the line of action and is characterised by a mixed sliding and rolling motion. At the pitch point, there is purely rolling motion although outside the pitch point, which can be seen in g 2, there is always sliding. As the gear contact passes the pitch point, the direction of siding changes, thus creating time-varying forces. Friction has been reported to have an eect on both the DTE and noise of a gear pair [20]. Velex and Cahouet [21] compares a dynamic model which includes sliding friction with measurement result and show an important contribution from friction for the dynamic bearing forces for speeds below 200rad/s (1900rpm) for the investigated gear pair. As the speed increases further, the inuence of friction decreases rapidly. Rebbechi, Oswald and Townsend [22] performs measurements of dynamic tooth friction in order to calculate a friction coecient. The friction coecient reported is approximately 0.04 to 0.06 with increasing values at lower speeds. Many dynamic models has been presented in order to calculate the inuence of friction. [20, 23, 24]. S. He and R. Singh [25] reports that friction has marginal eects on the dynamic transmission error for helical gears, as compared with spur gears where the eect is much larger.

2.3

Shuttling forces

Shuttling forces were examined by Borner and Houser [11], and is the result of the motion of the contact lines across the gear surface as the gears mesh. As the contact lines move across the teeth, the resulting force is not centred on the tooth but moves slightly back and forth along the width of the gear. This results in time varying bearing 4

Ip Rbp

cm TEs Rbg

km

Ig

g
Figure 3: Singel degree of freedom gear model

forces which can be calculated in a quasi-static analysis. There has been little research covering the importance of the shuttling force although it is naturally included in models that calculate the dynamic contact conditions and do not rely on an excitation such as transmission error [21, 26]. Velex and Cahouet [21] notices dierences in the bearing reaction forces of the pinion and suggest the dierence is due to the shuttling force.

Dynamic gear modelling

Calculating static transmission error is a standard procedure, commercial gear calculation softwares are available to predict TE. As the rotational speed of the gears increase, TE changes from being a purely geometry and load dependant property to also including the dynamics of the system. As the mesh frequency and the harmonics coincide with dierent eigenfrequencies of the system, the resulting DTE is a function the dynamic properties of the system. Since the DTE describes the complete system , including gears, shafts, bearings and housing, this makes the DTE more complex to analyse. DTE is often predicted by dynamic models of gear systems. Static TE is commonly used as excitation in order to calculate bearing forces, enabling the prediction of vibration and sound radiation from the housing. To make a dynamic model of a gear system it is important to have a correct model of the gear-mesh interaction. Models vary from simple one-degree-of-freedom lumped models [27, 24, 28], which can be seen in g. 3, to large FE-models [29, 9, 14, 30]. The single-degree of-freedom model seen in g. 3 describes 5

the system as a purely rotational model without shafts. The system consists of two inertias (I) connected with the mesh stiness (km ) and mesh damping (cm ) at the base radius (Rb ) for pinion(p) and gear(g). The transmission error is treated as an angular displacement in the gear mesh. Despite the large range of complexity between published models, the aim is to obtain a model that describes the gear system accurately enough, depending on the properties of interest. Many models do not take the eect of axial forces, due to the helix angle or bending vibrations of the shafts, into account when modelling the gears. Since a gear system is a very complex system, many assumptions are made to simplify the system. Many dynamic gear models use TE as excitation and if the purpose of the model is to simulate rattle, no load TE can be used, since rattle in most cases occurs in lightly loaded gears [14]. It has been shown that gear with low contact rations(typically spur gears), loss of contact can occur for higher torques at the gear mesh resonance. Instead of using TE, the variation in mesh stiness is sometimes treated as an excitation [31], but in most cases the variations in mesh stiness is included in the TE. Gear whine models often use measured or calculated quasi-static TE under load as the excitation [32]. Velex and Ajmi [33] use a theoretical approach to determine the possibility of using TE as excitation when modelling gears. Their conclusion is that the major excitation is the dierence between the no load transmission error and the quasi-static transmission error under load. Previous research has often used quasi-static TE under load alone as the excitation. Limitations of the use of TE as the excitation are also discussed, such as gear body deections when the contact conditions between the teeth no longer can be considered to be quasi-static. In paper E, a dynamic model of a single gear pair is presented. The model is a torsional model with exible bearings including transmission error and shuttling forces. The shaft only has a torsional stiness and is considered to have an innite bending stiness. The rotation is modelled as a reference frame with superimposed exible degrees of freedom. Gear mesh forces and bearing forces are examined for two dierent gear designs.

Measurement of gear noise and DTE

A large number of papers concerning gear dynamics have focused on dynamic models. Little work concerning the comparison between simulation and measurements has been published. Research on multi-stage gearboxes are especially hard to nd. Special equipment has to be used when TE is measured on a complete gearbox. Instead of just giving information of the shape deviations of the gear anks, the measurement describes the complete gear system. Factors such as shape deviations, misalignment of shafts, eccentricities and clearances in bearings all have inuence on the results. Deections also have inuence if the system is measured under load. Since many gearboxes consist of more than one gear pair, the measurements includes transmission errors from several gear pairs. A complete gearbox is a very complex system and the result can be very dierent 6

kerblom and P when disassembling and reassembling the gearbox. A arssinen [16] reports up to 7dB dierence in overall noise level for certain speeds. Akerblom also reports large dierences in noise levels depending on bearing pre-load. In paper D, the noise level from two dierent gear pairs were tested. The test object was a two-stage gearbox where the investigated gears were located as the rst gear stage. The noise level from the second gear stage, which was the same during all the test, was examined in order to investigate the dynamic properties when the rst gear stage has been changed. The two dierent measurements were surprisingly similar, indicating that the dynamic properties of the system remained the same when the rst gear stage was changed. The denition of TE and DTE also becomes more complicated for a complete gear system. Often encoders are used and the location of the encoders aects the result. Similar dierence can be seen when comparing single-degree-of-freedom (SDOF) models and multi-degree-of-freedom (MDOF) models. In the case of a MDOF model, the results depend on which nodes that are used to calculate the TE. In most cases, the measurement determine the denition of TE since the location of the encoders can be dicult to change. Measurements of dynamic transmission error have been performed in order to investigate non-linear behaviour of lightly loaded gears. The measurement setup is often designed in order to model the gear pair as a single degree of freedom system[14, 15]. The excitation used is transmission error and a good correlation with measurement result has been obtained. For paper A to C, DTE and noise were measured on a truck gearbox in a noise test rig. The gearbox is powered by an electrical motor, capable of speeds up to 2000 rpm. Another electrical motor is used for torque reaction. Each motor is located in a separate room next to the measurement room, in order to minimise the noise contribution. Walls and ceiling are covered with acoustic absorbents to reduce the acoustic reections. Rotational encoders are located on the input shaft, side shaft and output shaft. The encoder location can be seen in gure 4. This enables calculation of the DTE from the rst and second gear stage separately, as well as for the complete gear system. Four microphones and four accelerometers were used. Measurements were made at dierent torques and speeds, typical of the operating conditions of the gearbox. The measurement object was a standard gearbox from Scania ,similar to the gearbox displayed in g. 1. The gearbox has a three-speed main section, a range and a splitter function, resulting in 12 forward speeds. The range gear is a planetary gear with two possible gear ratios. In all measurements, the planetary gear was set in high range, which means that the complete planetary gear rotates as one unit. Therefore, the analysis does not include the planetary gear. Instead the properties of the helical gears in the main section of the gearbox are investigated. All gears in the gearbox are rotating and contribute to the total noise level. The noise from the unloaded gears is of broad band character and no gear mesh orders for the unloaded gears can be detected. The main gearbox has three gears which can be used in high or low split, resulting in six gears. The torque is transmitted from the input shaft to the side shaft via the high or 7

Range planetary gear

3x encoder

1:st gear mesh

2:nd gear mesh

Figure 4: Schematic view of the gearbox

low split gear pair. From the side shaft to the output shaft, the gear pair corresponding to the rst, second or third gear is selected and locked to the output shaft. In g. 4 a sketch of the working principles of the gearbox can be seen with the location of the encoders included. For paper D the measurement facility is the same and a similar measurement object is used, although twelve microphones were used instead of the previous four and no measurement was made of the rotation of the shafts. Two dierent gear designs were used in the measurements, details can be found in the paper.

5
5.1

Summary of main results

Paper A

Correlation between noise and DTE is examined. The method uses xed speed and calculates the correlation as the torque is increased. Eight dierent speeds, typical of the operating conditions of the gearbox, were used during the measurements. At each speed, the torque was varied between 600 to 2000 Nm in 200 Nm increments. The correlation between noise and DTE varies for dierent gear pairs. The highest correlation is found for the low split gear pair, which is located closest to the gearbox housing. The correlation also seems to increase as the applied torque increases. There can be a number of reasons for a reduction in the correlation between noise and DTE. Some resonances can be found where the correlation is low, but not all resonances result in low correlation. Comparing the measured DTE and calculated static TE for the gear teeth shows an increase in DTE and noise but a decrease in calculated TE as the torque is increased, 8

Figure 5: Calculated static p-p TE and measured dynamic p-p TE for three dierent speeds for the rst gear stage using the fth gear in low split

Figure 6: Calculated static TE and noise level (lp ) for three dierent speeds for the fth gear in low split

Figure 7: Averaged DTE of one tooth 1: 600 Nm; 2: 800; 3: 1000; 4: 1200; 5: 1400; 5: 1600; 7: 1800; 8: 2000Nm

as can be seen in g. 5. If the system is linear and the static TE is considered to be the excitation, the DTE should follow the same pattern. Non-linear eects and other sources can be an explanation to the lack of correlation between the calculated static TE and measured DTE. Similar conclusions are made when comparing the noise and the calculated TE, which can be seen in g. 6. The correlation between noise and DTE is much better. As the torque is increased the same trend can be seen in both DTE and noise.

5.2

Paper B

Dynamic transmission error has a close relation with the excitation of a gear system. Below a certain limit, it is independent of the speed of rotation but has a certain dependence on the torque applied, due to the limited stiness of teeth. An average of the DTE for a gear mesh can be seen in g. 7. The radiated sound from a gear system is generally increased with the increase of the rotating speed at the rate of 6 dB per doubled rotating speed, with the uctuations due to the shifted transfer functions, as can be seen in g. 8. The sound pressure is very dependent on the structure and how the gear pairs are mounted. Identical gear pairs may produce totally dierent sound power when mounted in dierent systems or stages. It is possible to estimate the transfer function between DTE and the sound pressure by using a speed sweep measurement.

10

Figure 8: Sound pressure levels at the meshing frequency of the rst gear stage The center dotted line shows 6 dB per doubled speed.

5.3

Paper C

Measurements of DTE and gear noise have been used to assess the possibilities of using a linear model to describe the system. An order tracking method using a Kalman lter was employed to calculate the amplitudes of the gear mesh orders. Close orders and overlapping sidebands limited the possibility of using a standard FFT order tracking method. The magnitude of the transfer function between DTE and noise is calculated using the result from the Kalman lter. There seems to be a linear relationship between noise and DTE for the higher torques measured, which is consistent with the result in paper A and paper B. Many dynamic models of gear systems use a linear approach to calculate dynamic forces and the radiated noise. Assuming a dynamic model that is linear with rotating speed, the non-linear eects in the measurement will be the error of the model. For each frequency, the average of the normalised gear mesh order and harmonics are calculated. No information can be given the rst part of the gear mesh order, since an overlapping order is required to assess the dierence between the orders. Therefore no information can be given below 1200 rpm for the gear mesh order. This results in an average transfer function for all the orders. To calculate the gear mesh order, the estimated transfer function is multiplied with a suitable constant, the same applies to the rest of the harmonics. 11

Without friction force (a=0)

1 0,8 1 0.8

With friction force (a=1)

Correlation coefficient

0,6 0,4 0,2 0 0,2 0,4 0,6 0,8 1 0 20 40 60 80 100

Correlation coefficient

0.6 0.4 0.2 0 0.2 0.4 0.6 0 20 40 Gear mesh order A First harmonic A Gear mesh order B First harmonic B 60 80 100

Stiffness k(kN/m)

Stiffness k(kN/m)

Figure 9: Correlation coecient between suggested excitation and measured noise level for dierent values of k

5.4

Paper D

Two gear pairs with dierent micro geometry but similar macro geometry has been tested using a complete truck gear box in a noise test rig. The dierence in noise level for dierent torques can not be explained using transmission error alone as excitation. The measurement result indicate that for this gearbox, shuttling forces and friction forces can be very important when assessing the noise properties of the tested gears. Minimising transmission error does not necessarily minimise noise even though the transmission error multiplied with the mesh stiness seems to be the dominating excitation force. The reason is that multiplying TE with the mesh stiness signicantly overestimates the importance of transmission error for a gear pair not operating above the critical gear mesh resonance. In a complex gear system, the deection does not necessarily take place at the gear mesh. The correlation coecient is calculated between the measured noise level and the suggested excitation for dierent values of stiness k , which is multiplied with the transmission error. The correlation coecient or gear mesh orders and rst harmonics for both gear pairs can be seen in gure 9, with and without the contribution from friction. Multiplying the transmission error with a much weaker stiness signicantly increases the correlation with the measured result, something which reduces the TE excitation to the same level as shuttling forces and friction forces.

5.5

Paper E

A dynamic model has been set up using transmission error and shuttling force as excitation. In the sub-harmonic region, the system behaves quasi-statically. If all forces

12

acted along the line of action and a constant torque was applied by an engine and break there would not be any uctuating gear mesh forces, only a constant gear mesh force due to the applied torque. Since the constant gear mesh force moves back and forth over the gear tooth there is a shuttling force at the bearings. In the region above the rst resonances, dominated by the engine and break, but below the next resonance, the time varying bearing force is controlled by the shaft stiness and the loaded transmission error together with the shuttling force. The reason is that the inertia of the gear and engine are vibrationally stationary, only rotating with a constant angular speed. Since the shafts in this model are considerably weaker than the gear mesh, which often is the case, the transmission error acts as a forced angular displacement. The weakest component, here the shafts, then deects which determines the reaction force in the gear mesh. Due to the small time varying force due to transmission error, shuttling forces can become very important and dominate the excitation in this frequency region. At super critical frequencies, above the highest resonance frequency, the system is controlled by the inertia forces. This means that the rotations of all inertias are a function of the rotating speed and gear ratio. Another way of describing the motion is to say that the peak-to-peak angular transmission error is zero between all inertias. The deection of the gear teeth is then equal to the loaded static TE and the force equal to the deection (T Es ) multiplied with the active stiness which is the gear mesh stiness, km . Some conclusions can also be drawn for the resonant areas. For the resonances which primarily includes the torsional degrees of freedom, the amplitude of the gear mesh force and bearing force is depending on the transmission error. Depending on the mode shape, dierent stinesses will be important. For the lower torsional resonances, the shafts stiness are the most important and the trend of the gear mesh force is that of the transmission error multiplied with a constant stiness. For the critical gear mesh resonance, the most important stiness is the gear mesh stiness and the trend of the gear mesh force therefore follows the trend of transmission error multiplied with the mesh stiness. The rocking mode of the shaft is excited by the shuttling force and therefore following the trend of the shuttling force. Although other sources of gear noise such as friction and bending moments has been proposed, TE has often been seen as the dominating excitation [11]. One of the reasons is that the TE excitation force has been approximated as km T E [11, 34], but below the gear mesh resonance the gear mesh force is dominated by the TE and the shaft stiness. Since the shaft stiness in this case is signicantly lower than the gear mesh stiens, the corresponding force will also be lower, increasing the importance of other sources such as shuttling forces.

Conclusions and future work

The measurements of dynamic transmission error and noise show a correlation between the two properties of the tested gearbox. The correlation was highest for the input gear stage located close to the gearbox housing and lower for the second gear stage, which 13

is located closer to the centre of the gearbox. One reason for the lack of correlation between DTE and noise can be that other forces such as shuttling and friction forces have little inuence on DTE but a larger inuence of emitted noise. This is evident for the measurements presented in paper D where the explanation for the dierence between the measured gear pairs can be found in the friction forces, but most of all the shuttling forces. This shows the importance of including other forces beside transmission error and also to have a dynamic model of the gear system in order to accurately predict the noise behaviour of a gear system. Since dierent excitations have a dierent sensitivity to dierent resonances of the system, a dynamic model of the system is needed. Much of the work done in gear dynamics has to do with gears without manufacturing errors. In the future, the inuence of production error such as runout, pitch error and gear shape errors must be included in the dynamic models in order to predict the dynamic excitation for a gear pair with typical manufacturing errors. Depending on the chosen manufacturing process, it would be of interest to be able to predict the standard deviation of bearing forces for a specic gear design and production method.

14

References
[1] Houser, D. R., 1986, The root of gear noise-transmission error, Power Transmission design, 86(5), pp. 2730. [2] Council Directive 70/157/EEC, 1970, Permissable sound level and the exhaust system of motor vehicles, . [3] Commission Directive 2007/34/EC, 2007, Amendment of directive 70/157/EEC, . [4] Welbourn, D. B., 1979, Fundamental knowledge of gear noise - a survey, Proc. Noise and vib. of Eng. and Trans., I Mech E, Craneeld, UK, pp. 914. [5] Townsend, D. P., 1992, Dudleys Gear Handbook, 2nd edition, McGraw-Hill Inc, ,page 14.5. [6] Mao, K., 2006, An approach for powertrain gear transmission error prediction using non-linear nite element method, Proc. IMechE Part D: Journal of Automobile Engineering, 220, pp. 14551463. [7] Maki, H., 2001, A study of tooth modications of helical gears relevant to gear noise, MPT2001-The JSME International Conference on Motion and Power Transmission, Nov. 15-17, Fukukoka, Japan, gDN-11. [8] Bonori, G., Barbieri, M., and Pellicano, F., 2008, Optimum prole modications of spur gears by means of genetic algorithms, Journal of Sound and Vibration, 313(3-5), pp. 603 616. [9] Lundvall, O., 2004, Contact Mechanics and Noise in Gears, Ph.D. thesis, Department of mechanical engineering, Link oping University. [10] Houser, D. R., Vaishaya, M., and Sorenson, J. D., 2001, Vibro-acoustic eects of friction in gears: An experimental investigation, SAE technical paper series, 2001-01-1516. [11] Borner, J. and Houser, D. R., 1996, Friction and bending moment as gear noise excitation, SAE technical paper series, 961816. [12] Davies, G., Brooks, P., and Findlay, M., 2001, Recent advances in automotive gear pair dynamic behaviour measurement and prediction - a review, The JSME International Conference on Motion and Power Transmissions, pp. 9096. [13] Houser, D. R. and Singh, R., 2002, Gear Noise Short Course, Department of mechanical engineering Ohio state university Columbus.

15

[14] Parker, R. G., Vijayakar, S. M., and Imajo, T., 2000, Non-linear dynamic responce of a spur gear pair: Modelling and experimental comparisons, Journal of sound and vibration, 237(3), pp. 435455. [15] Al-shyyab, A. and Kahraman, A., 2005, Non-linear dynamic analysis of a multimesh gear train using multi-term harmonic balance method: period-one motions, Journal of Sound and Vibration, 284, pp. 151172. kerblom, M., 2008, Gearbox Noise, Correlation with Transmission Error and In[16] A uence of Gearing Preload, Ph.D. thesis, Department of Machine Design, Royal Institute of Technology, tRITA-MMK 2008:19, ISSN 1400-1179. [17] Houser, D. R., Oswald, F. B., Valco, M. J., Drago, R. J., and Lenski, J., 1994, Comparison of transmission error predictions with noise measurements for several spur and helical gears, 30th AIAA/ASME/SAE/ASEE joint propulsion conference, Indianapolis. [18] Yildirim, N., Gasparini, G., and Sartori, S., 2008, An improvement on helicopter transmission performance through use of high contact ratio spur gears with suitable prole modication design, Proc. IMechE Part G: Journal of Aerospace Enginering, 222, pp. 11931210. [19] Opitz, H., 1968, Noise of gears, Philosiphical Transactions of the Royal Society of London, Series A, Mathematical an Phyical Sciences, 263(1142), pp. 369.380. [20] Lundvall, O., Str omberg, N., and Klarbring, A., 2004, A exible multi-body approach for frictional contact in spur gears, Journal of sound and vibration, 29(4), pp. 100200. [21] Velex, P. and Cahouet, V., 2000, Experimental and numerical investigation on the inuence of tooth friction in spur and helical gear dynamics, Journal of Mechanical Design, 122(1), pp. 515522. [22] Rebbechi, B. and Townsend, F. B. O. D. P., 1996, Measurement of gear tooth dynamic friction, NASA Army Research Laboratory, Technical Report ARL-TR1165. [23] Vedmar, L. and Andersson, A., 2003, A method to determine dynamic loads on spur gear teeth and on bearings, Journal of sound and vibration, 267(3), pp. 10651084. [24] Vaisha, M. and Singh, R., 2001, Analysis of periodically varying gear mesh systemwith coloumb friction using oquet theory, Journal of sound and vibration, 243(3), pp. 525545.

16

[25] He, S. and Singh, R., 2008, Dynamic transmission error prediction of helical gear pair under sliding friction using oquet theory, Journal of mechanical design, 130, pp. 052603 19. [26] Eritenel, T. and Parker, R. G., 2009, Computational nonlinear vibration analysis of gear pairs using a three-dimensional model, IDETC/CIE 2009, pp. paper DETC200987485. [27] Amabili, M. and Fregolent, A., 1998, A method to identify modal parameters and gear errors by vibrations of a spur gear pair, Journal of sound and vibration, 214(2), pp. 339357. [28] Wang, Y. and Zhang, W. J., 1998, Stochastic vibration model of gear transmission error considering speed dependant random error, Nonlinear dynamics, 17, pp. 187 203. [29] Miayauchi, Y., Fujii, K., Nishino, T., Hatamura, K., and Kurisu, T., 2001, Introduction of gear noise reduction ring by mechanism analysis including FEM dynamic tuning, SAE techical paper serise, 2001-01-0865. [30] Celik, M., 1999, Comparison of three teeth and whole body models in spur gear analysis, Mechanism and machine theory, 34, pp. 12271235. [31] Theodossiades, S. and Natsiavas, S., 2000, Non-linear dynamics of gear-pair systems with periodic stiness and backlash, Journal of Sound and Vibration, 229(2), pp. 287310. [32] Karaman, A. and Singh, R., 1991, Non-linear dynamics of a geared rotor-bearing system with multiple clearances, Journal of sound and vibration, 144(3), pp. 469 506. [33] Velex, P. and Ajmi, M., 2006, On the modelling of excitations in geared systems by transmission errors, Journal of Sound and Vibration, 290(2), pp. 882909. [34] Houser, D. R. and Harianto, J., 2008, Microgeometry and bias in helical gear noise excitation, Gear Solution, Februari, pp. 2139.

17