Guoxing Li - TEPEN

Taiyuan University of Technology

University of Huddersfield

24th Oct 2020

Guoxing Li

Background & significance Modelling Fault diagnosis Conclusions

Background and significance

Modelling of tribo-dynamic behaviours

Applications in mechanical fault diagnosis

Conclusion and future work

Outline

Motivation and Significance

Tribo-dynamic behaviours widely exist in variousmechanical engineering fields.

Tribo-dynamics is a new field that has emergedfrom the confluence of structural dynamics,contact mechanics and tribology. Central tothese three fields is the study of interactionsbetween interfaces; the difference comes in thelength scale considered and the tools used tostudy an interface interaction.

Modelling of tribo-dynamics between interactiveinterfaces is of great significance for subsequentmechanical fault diagnosis and reliabilityoptimization.


Motivation and Significance

Deterioration of tribo-dynamics behaviour can causeserious failures and losses. Therefore, it is of greatimportance to implement an accurate conditionmonitoring of early faults in friction pairs.

Prognostic of fatal accidents and estimation of residual life.

Guidance for optimization of operation and maintenance.

Guidance for optimal design of friction pairs to improve efficiency and reliability.

……


Outline







It is necessary to combine structural dynamics, tribology and contact mechanics to model tribo-dynamic behaviours in multi-disciplinary and cross-scale. According to the length scale of interface interaction, models can be divided into three types

(Cross-scale modeling)

e.g. Transient elastic waves (Acoustic Emission) produced by micro-tribological behaviours

Macroscopic model

Combination of dynamic model and tribological model

Mesoscopic model Influences of micro characteristics on macro response

Microscopic modelWeak dynamic response caused by micro-interface interaction


Macroscopic model of tribo-dynamics

The continuously knocking behaviour of piston against the cylinder liner driven by gasand inertial force is a typical tribo-dynamic behaviour. It includes impact mechanics,structural transient dynamics, elasto-hydrodynamic lubrication, boundary lubrication, etc.

Combination of dynamic and tribological model



Combining structural dynamics and elastohydrodynamics to construct a fluid-solid coupling model of piston assembly is helpful to reveal the coupling relationship between in-cylinder combustion behaviour, structural dynamics and physio-chemical properties of lubricant.

Combination of dynamic and tribological model


1 2 3 4 50

2000

4000

6000

8000

1 2 3 4 50

5

10

Measured vibration

Simulated vibration

The tribo-dynamics model can effectively predict the non-linear trend of vibration response with increasing speed.


RMS of liner vibration

slap-induced vibration after combustion TDC


A B CE F D

Mesoscopic model of tribo-dynamics

Generally, the length scale of structural dynamic behaviour is on the order of micrometer-millimeter, and the scale of geometric interface characteristics is on the order of nanometer-submicron. To study the interaction between cross-scale factors, such as the effect of surface roughness on structural vibration, it is necessary to merge the two on meso-scale.

Case 1: Introduction of high-frequency surface deformation into lubrication modeling

Case 2: Vibration characteristics induced by tribofilm-asperity interactions in hydrodynamic journal bearings


Introduction of surface deformation into lubrication model

Establishment of Finite Element Model

(Modal characteristics) Structural transient dynamics simulation

(Contact constraints) Nonlinearities of assembly constraint

Case 1


(a) Transient dynamic simulation in ANSYS workbench

(b) Deformation result imported into post-processing software

(c) Node displacements in Cartesian coordinate

(d) Node displacements in a 2D coordinate system containing the axial and circumferential directions.

(e) Filtered deformation

(f) Improved oil film thickness distribution

Case 1Introduction of surface deformation into lubrication model


The simulated friction force based on the improved model shows:1, modal-related friction fluctuations2, reduction in viscous friction ↘3, sharp peak of asperity friction around 380° ↗

Case 1


Introduction of surface deformation into lubrication model

Vibration-induced friction reduction

Vibration induced by tribofilm-asperity interactions

3 2 26 12p h hh R Rt

Infinitely long bearing theory

3 2 26 12o o oo

p h hh R Rt

3 2 2 26 12 3asp asp asp oo o asp

p h h ph R R h ht

Decoupled perturbation Reynoldso asph h h

o aspp p p

ho=c[1+εcos(θ-Φ)]

According to the perturbation theory, the bearing will undergo small amplitude oscillations at its steadystateposition which may caused by interaction between tribofilm-asperities.

Since h0 and hasp are of the order of millimetre and micrometre respectively, hasp can be reasonably viewed as a perturbation component to the macroscopic film thickness h0.

Case 2


Journal bearing test rig

Journal bearing test rig and sensors location

Self-aligning journal bearing components

Case 2

Surface topography of the journal on shaft 3:

(a) photomicrograph (Ra=2.51 µm) (b) 3D surface profile

(b) (c) height probability distribution (d) power spectral density.


Vibration induced by tribofilm-asperity interactions

Vibration spectrum with the worn journal surface under different operating conditions:

radial load of 3 bar

Case 2


As the journal surfaces topography changed from mild wear to severe wear, the amplitudes of the spatial components increase particularly at the low spatial frequencies. Therefore, it generates a strong excitation within a certain frequency range. With the increase of the speeds (1200rpm-2100rpm), the frequency bandwidth of resonant responses becomes wider and moves higher.

1. The random pressure fluctuation occurs between the tribofilm and dynamic asperities is one of the significant excitation sources for the resonant vibrations of journal bearings in the hydrodynamic lubrication regime.

2. The excitation intensity is tightly related to the journal surface parameters and operating conditions.

Microscopic model of tribo-dynamics

Acoustic Emission (AE) describes the radiating phenomenon of transient elasticwaves caused by the redistribution of stress in the material. Generally, AE canbe generated from various sources, such as fluid friction, delamination, crackgrowth and sliding friction. The released energy from these activities has a lowamplitude and wide frequency band.

Establishing a model of AE generation from tribo-dynamics and developingweak feature extraction algorithm is very important for non-intrusive monitoringof early friction and wear in rotating machinery and reciprocating machinery.

Case 1: Modelling acoustic emissions generated by asperity-asperity contact(Yibo Fan, Fengshou Gu, Andrew Ball, 2010)

Case 2: Modelling acoustic emissions induced by dynamic fluid-asperityshearing (Jiaojiao Ma, Hao Zhang, Zhanqun Shi, Fulei Chu, Fengshou Gu,Andrew D. Ball, 2020)


Acoustic emissions generated by asperity-asperity contact

A theoretical model based on the Greenwood-Williamson(GW) model has been developed to validate therelationship between AE pulses and the elastic strainenergy generated by the asperity-asperity contact on theroughness surfaces.

• Fan Y, Gu F, Ball AD. Modelling acoustic emissions generated by sliding friction. Wear 2010; 268: 811–5.

Asperity contacts on rough surfaces

312 24( )

3e eL ER

12

12

1_

(h)0.3(h)AE asp c e

FU K NL vRF

12

12

1(h)0.3(h)E e

FU NL vRF

is release rate of elastic energyKc is portion of the elastic strain energy converts to AE pulses Le is the elastic loadFn(h) is probability density function

The level of AE measurement is determined by the load supported by asperity contact, sliding speed, the number of asperity contact and surface topographic characteristics.

EU


Acoustic emissions generated by fluid-asperity shearing

Nevertheless, the mechanism and time-frequency characteristics of AE signals generated from the process of dynamic fluid-asperity shearing (FAS) are still ambiguous in the local level.

In hydrodynamic lubrication regime, where there are no micro-asperity contacts, as shown on the right, the asperity deformation is nil.

Changes in shear force will induce different deformations to the asperities. As the deformations of asperities recover in a short time duration, the transient rapid release of energy transfers to the plate as a form of AE waves.

Low stress Low stressHigh stress

Asperity-valley Asperity-asperityAsperity deformation

Asperity-valleyAsperity recovery

u

(a)

asp1

- asp2

h0h

FF F

F

F F

h1 h2 h3

u

(b)

u

(c)

FF F

F

F F

AE waves


Fluid-asperity shearing (FAS) in lubricated rough surfaces

Acoustic emissions generated by fluid-asperity shearing

ℎ 𝑥,𝑦, 𝑡 ℎ 𝛿 𝑥,𝑦, 𝑡

𝜏 𝑥,𝑦, 𝑡 𝜂 e𝑢

ℎ 𝛿 𝑥,𝑦, 𝑡ℎ

2 ℎ 𝛿 𝑥,𝑦, 𝑡

𝜕𝑝𝜕𝑥

The combining amplitude of asperities on the relative sliding surfaces is

𝛿 𝑥,𝑦, 𝑡 𝛿 𝑥 𝑢𝑡,𝑦 𝛿 𝑥 𝑢𝑡,𝑦

The instantaneous fluid film thickness isthe instantaneous film thickness between the relative sliding surfaces is a function of time/space.

Therefore, the dynamic FAS model is derived as

Due to the consideration of asperity topographic characteristics, oil physic-chemical properties and relative movement parameters, FAS model can characterize the acoustic emission level caused by fluid-asperity shearing.


Outline







Macroscopic model

Mesoscopic model

Microscopic model


For different types of tribo-dynamic behaviour and failure events, based on models of different length scales, a variety of condition monitoring and fault diagnosis technologies have been developed.

Vibration-based

AE-based

Dynamic behaviours

other events related to dynamics ...

Tribological behaviours

other events related to tribology ...

Vibration-based lubrication monitoring of piston-cylinder pairs

Vibration-based combustion status monitoring for IC engines

Vibration-based NOx/PM emission monitoring for engines

AE-based elastohydrodynamic lubrication monitoring

AE-based lubrication monitoring of IC engines

AE-based abnormal wear detection for IC engines

AE-based gear wear monitoring

AE-based prognostic of wear and scuffing failures in IC engines

Vibration-based lubrication monitoring of piston-liner pairs

Test rig for cylinder liner vibration measurement

DAQ

Dynamometer

Acc

eler

omet

er

Pressure Sensor

Ang

ular

spee

d

Control system



Piston SlapCombustion

Com. TDCExh. TDC

1st modal

4th modal

TS ATS

CWT of measured vibration

Based on macroscopic tribo-dynamic model, various excitation sources associated with different local vibration responses can be corresponded and monitored.




TS ATS

With the increase of oil viscosity, the magnitude of the impact response basically shows a downward trend.

Experimental results show that differences in the lubricant condition can definitely cause observable changes in the dynamic responses of cylinder liners in IC engines.

0°/360°

180°

90°270°

TDC720°

540°BDC

E

FA

B

C

D

I

G

450°630°

Hw

avel

et c

oeffi

cien

t

wav

elet

coe

ffici

ent

(a) 10Nm (b) 40Nm.

RMSs of slap responses after exhaust TDC

Vibration-based combustion status monitoring

CWTs of vibration measured in engines burning different fuels

Diesel Biodiesel


Vibration-based combustion status monitoring

RMS of vibration response after combustion TDC fueled withbiodiesel shows a nonlinear tendency of decrease after an increase with the increase of speed

Δt

2sin / 1 ( sin )=( )y c i c iF F F t tp p

The advanced ignition caused by the high cetane number of biodiesel has a significant contribution to the compound effect on the piston side-thrust force, thereby resulting in a nonlinear trend of the RMSs of local response near the combustion TDC.



vibration combustion exhaust emission

reconstruct evaluate

Vibration Based Virtual Sensing of NOx/PM Emission in CI Engines



c.Structural similarity index (SSIM) analysis

d.Morphological reconstruction

a.Selection of source data

b.Image graying


1 : Measured emissions；2-12: Predictor variables.

Operatingconditions

Am

plit

ude

[-]

a. Predictor variablesb. Principal component analysis


There is a good correlation between structural vibration characteristics and measured NOx emissions. This provides a novel way for low-cost transient emission monitoring.

c. Comparison of predicted and measured NOx



Fig. 4. Cone-plate based a rheometer

𝜏 𝑟,𝜃, 𝑡 𝜂 eΩ 𝛼⁄

1 𝛿 𝑟,𝜃, 𝑡 𝑟𝛼⁄

where 𝛿 𝑟,𝜃, 𝑡 𝑧 𝑠𝑖𝑛 2𝜋𝑟𝜒 𝜃 𝛺𝑡 𝜁

Ω

rR

z

0

rx

z

yθ

rp

α

h0

Dynamic FAS resulting from the cone-plate with the roughness surface:

𝑎𝑛𝑑 𝑧 2𝐺 𝜒 , Δ𝜒

Ω : angular speed 𝜃 : the angle in the range from 0 to 2𝜋 degree.𝜁 : random initial phase angles. 𝐺 : the spatial PSD of the random heights𝜒 : the spatial frequency𝜒 , : the centre frequency Δ𝜒 : the width of each frequency interval

: shear rate: radius

h : radius

𝛼 𝑠𝑖𝑛𝛼 𝑡𝑎𝑛𝛼

ℎ 𝑟𝑡𝑎𝑛𝛼 𝑟𝛼

𝛾𝛺𝑟ℎ

𝛺𝑟𝑟𝛼

𝛺𝛼

𝛾𝑟

Rheometer test rig and AE sensor location

Fluid-Asperities Shearing (FAS) model is derived to investigate the influence of surface profiles, lubricants and operating conditions on AE characteristics. Experiments are carried out based on a rheometer rig.



Results and discussion Results and discussion Results and discussion

Characteristics of the dynamic FAS simulated and AE measured with different operating conditions: (a) normalised RMS and (b) highest Frequency of the frequency band.

It has been found that the correlation length of surface roughness parameters and shear rate are

two main factors affecting FAS behaviours and consequently AE characteristics.

It shows that AE signals can effectively characterize the changes in surface topography

parameters and shearing rate in HL regime.



Direction of piston motion

In-cylinderpressure

piston

Line

r

asperities

ringliner

film

12

12

1_

(h)0.3(h)AE asp c e

FU K NL vRF

EU

The level of AE measurement is determined by the load supported by asperity contact, sliding speed, the number of asperity contact and surface topographic characteristics.

Oil film

piston

Direction of piston motion

Line

r

ring

asperities

Asperity–Asperity Contact Fluid-Asperity Shearing

0

( ) ( ) ( , , ) ( , , ) ( , , )( ) [ ( , , )]3 2 ( , , )

izr b p d ib

l i

d k v t N t z w h x y t p x y t x y tAE t v x y t dyE I y h x y t




Schematic diagram of the diesel engine test system

AE WPT spectra in power stroke for the used oil at different speeds and high load.

Diversities of changes in AE amplitudes in different bands can be effective indicators of the degradation of lubricating oils, as the deviations from linear trends can provide additional reliable information for differentiating between baseline oils and used oils.


Wear detection of piston components

Acoustic emission and vibration sensors are installed on the surface of cylinder block near the combustion chamber to detect abnormal collisions and friction events pf piston assembly.

Acoustic emission and vibration signals measured on engine block

0 90 180 270 360 450 540 630 7200

0.5

1

E

VO

E

VC

IV

O

IV

C

Power

幅值

(V)

0 90 180 270 360 450 540 630 7200

20

40

E

VO E

VC

IV

O

IV

C

Power

曲角轴转 (°)

幅值

(V)

piston

Piston pin end


Ampl

itude

（V）

Ampl

itude

（V）

Angle（°）

An abnormal wear on the small end of a connecting rod was found.


Fault diagnosis of bearing bush

Combining the crankshaft displacement measured by the laser sensor and the acoustic emission signal, the radial force of the bearing bush was calculated, and then an early wear failure of bearing bush was diagnosed.

0 90 180 270 360 450 540 630 720-2

0

6x 104 直方向的承力竖轴

曲角轴转 (o)

承力

轴(N

m)

10Nm-1000r/min10Nm-1400r/min10Nm-1800r/min

0 90 180 270 360 450 540 630 720-2

0

2x 104

曲角轴转 (o)

承力

轴(N

m)

水平方向的承力轴

10Nm-1000r/min10Nm-1400r/min10Nm-1800r/min

10000Nm

30000Nm

50000Nm

90°

270°

180° 0°

Bearing forces in vertical and horizontal directions

Bearing loadBackground & significance Modelling Fault diagnosis Conclusions

Ampl

itude

（V）

Load

(N)


The position where the abnormal acoustic emission signal occurs is exactly the same as the timing of the peak of the oil pressure change rate curve (that is, when the pressure relief valve is opened), indicating that the pressure relief valve is working abnormally.

Sensor layout and lubrication circuit

0 90 180 270 360 450 540 630 7200

1

2

3

4

5

6

7

曲角轴转 (°)

幅值

(V)

限压阀异常声射信号发比对图- 1800r/min-10Nm

油前调节压

油后调节压

0 90 180 270 3600.65

0.7

0.75

0.8

0.85

0.9

机油

力幅

压值

(MP

a),

0 90 180 270 360 450-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

异常声

射信号

幅发

值 (V

)

0 90 180 270 360-200

-100

0

100

200

300

机油

力化率幅

压变

值 (M

Pa/

s)

曲角轴转 (°)

机油力压

机油力化率压变

异常声射信号发

Oil pressure, pressure change rate and abnormal acoustic emission signal

Fault diagnosis of pressure limiting valve Normal and abnormal acoustic emission signals


Ampl

itude

（V）

AbnormalNormal

Oil

pres

sure

（M

Pa）

Oil presuurePressure change rateAE signal

Pres

sure

cha

nge

rate

（M

Pa）

Ampl

itude

（V）

AE-based prognostic of wear and score failures

AE corresponding to different oil supply

piston seizure/ cylinder score

Extract characteristic signals related to asperity-asperity contact events from measured acoustic emission signal, to monitor the occurrence of abnormal wear in real time, and to provide early warning of cylinder failure.


As the supply of lubricating oil decreases, the amplitude of the acoustic emission signal increases significantly.

AE-based gear wear monitoring

0°: Boundary lubrication 0°-2.9°: Mixture lubrication 2.9°: Boundary lubrication 4.1°： close to 0

4.1°- 5.2°: Boundary lubrication 5.2° : Hydrodynamic lubrication 9.4°: speed increase, mixture lubrication

Ampl

itude

angle

peak

viscosity

peak

peak

speed /% load /%

loadloadloadload

loadloadloadload


Outline






Conclusions

Tribo-dynamic behaviours can be modeled in combination with structural dynamics, tribology, and contact mechanics. According to the length scale of interface interaction, models of tribo-dynamics can be divided into three types: Macroscopic model, Mesoscopic model and Microscopic model.

According to the perturbation theory, macro dynamic response can be affected by micro tribo-dynamic interactions, which provides a new way to achieve condition monitoring of cross-scale behaviour and early fault diagnosis.

For different types of tribo-dynamic behaviour and failure events, based on models of different length scales, corresponding technologies can be developed for monitoring and diagnosis.


Technology Outlook

Condition monitoring and failure warning of high value-added mechanical systems

The potential of cross-scale feature detection demonstrated by perturbation theory provides novel approaches for early fault diagnosis and residual life prediction.

Based on in-depth theoretical understanding, the optimal design of friction pairs can be further carried out to pursue friction reduction and reliability improvement.


Guoxing Li

Taiyuan University of Technology

Tel: +86 13233687325Email: [email protected]

Documents

Guoxing Li - TEPEN