Wet Sliding Wear Behavior of EN31 Rolling Contact Bearings for Screw Compressor Application with Statistical Analysis

The aim of this work is to estimate the wear rate of EN31 material & obtain the line of best fit using statistical analysis of wear. In industrial screw compressors, EN31 is used as a material of construction for cylindrical roller bearings. The wet sliding phenomenon between the outer surface of the inner race of the rolling contact bearing & the outer surface of the cylindrical roller is simulated using pin-on-disk tribometer. Measurement of wear, friction force & friction coefficient is carried out by rubbing different cylinders of hardened EN31 against hardened EN31 disk of the pin-on-disk tribometer. Total 25 tests each of 10 minutes time duration are performed to estimate wear with variable loads in the range of 20 N to 100 N and speed in the range of 692 RPM to 1246 RPM. From experimental investigation using pin-on-disk tribometer & statistical analysis using MINITAB © commercial software, it is observed that the material wear rate of EN31 increases with the increase in load as well as increase in speed. From pin-on-disk experimentation & statistical analysis it is concluded that the load factor has higher effect on wear than the speed factor. It was observed that the wear increases with the load as well as with the speed.


INTRODUCTION
The objective of this research work is to study the tribological behavior of the bearing steels used for construction of cylindrical roller bearings which form an important part of the industrial screw compressor system.Research in tribology is important for the reduction of abrasive wear & friction forces in any two mating machine components in relative motion.EN31 is the widely used material for construction of inner & outer races of rolling contact bearings and also the cylindrical rollers.During operation, there is a relative motion between the outer surface of the rotating inner race & the outer surface of the cylindrical rollers which generates friction at the interface & also results in abrasive wear of both the mating components.The aim of tribological investigations of any mechanical system is minimization of friction forces & abrasive wear by application of a proper lubricant between the two surfaces in contact.This section summarizes the research work carried out by various researchers in the field of cylindrical roller bearings to study the tribological behavior of the system in order to minimize the resulting friction & wear.
Harris, et al. [1] carried out research in rolling bearing technology and studied kinematics, contact mechanics & tribological aspects of the rolling bearing technology.They described in detail the kinematics of cylindrical roller bearings & derived the mathematical expressions for angular & linear velocities of the same.Beswick, et al. [2] worked on various steel materials used for construction of bearings and discussed the manufacturing technology along with load-speedlubrication aspects in the field of bearing technology.Bhushan [3] did fundamental research in the field of tribology to study experimentally the mechanisms of adhesive wear & abrasive wear for quantitative determination of material loss during the process of wear.Bhandari [4] explained the design of rolling contact bearings taking into consideration the industrial requirements along with selection from the vendor catalogues.Guillermo, et al. [5] carried out research on rolling bearing technology for refrigerant compressors and studied design of various types of refrigerant compressors and the influence of various lubricants on the tribological issues of rolling contact bearings.Li, et al. [6] evaluated cyclic contact stresses in cylindrical roller bearings using explicit finite element analysis and simulated fatigue crack initiation, crack growth & spalling.Liu, et al. [7] established a new method for modeling effects of natural defects on roller bearings and proposed a contact model taking into consideration the geometric features of the natural defects.Panda, et al. [8] maximized fatigue life of rolling contact bearing using Particle Swarm Optimization algorithm, whereas, Feldermann, et al. [9] used computational fluid dynamics to determine hydraulic losses in radial load cylindrical roller bearings.Denni, et al. [10] developed a novel multi-body dynamics model for simulation of rolling contact bearing mechanism in planetary applications.Laithy, et al. [11] investigated the effect of operating conditions such as pressure, Guo, et al. [13] developed models to estimate cage speed and roller speed of cylindrical roller bearings considering the roller centrifugal forces, friction forces at the contact region, contact elasticity between the roller & the raceway & hydrodynamic lubrication.Ahmer, et al. [14] used pin-on-disk tribometer to measure friction coefficient & wear rate of aluminum metal at room temperature.Stratmann, et al. [15] evaluated growth of tribofilm in case of cylindrical roller thrust bearings operated in boundary lubrication regime.El-Thalji, et al. [16] worked on dynamic modeling of wear evolution considering tribological & topographical changes.They used multiple force diagrams to simulate the dynamic input.Takabi, et al. [17] studied dynamic behavior of radially-loaded roller bearings operating at low as well as high speeds considering different values of surface roughness.Eyre [18] explained industrial significance of different types of wear such as abrasive wear, adhesive wear, erosion, corrosion & fretting of various metals.Zhang, et al. [19] carried out mathematical analysis of friction mechanisms in sliding-rolling contacts in presence of elasto-hydrodynamic lubrication considering asperity interaction friction induced due to solid-to-solid ploughing & boundary film friction.This development was found to be a cost-effective method of estimation of friction coefficient in mixed elasto-hydrodynamic lubrication.
Gawande et al. [20] simulated the phenomenon of sliding gear tooth in planetary gear train application of CNC bending machine using pin-on-disk apparatus with 5 minutes duration tests for 50 N & 100 N loads at 350 rpm disk speed to study the variation of wear, friction force & friction coefficient with time.Dhanvij et al. [21] carried out tribological investigations on EN24 material for evaluation of friction coefficient, friction force & wear rate by sliding it against EN31 disk having 60 HRC hardness.Nicholas, et al. [22] used novel ultrasonic reflectometry techniques for measurement of roller loads in high-speed wind turbine gearbox.They bonded piezoelectric transducers on the outer raceway of the bearing.From the reflected pulses, bearing load was determined.Cavacece, et al. [23] worked on life evolution of rolling contact bearings subjected to high local contact pressures & low oscillatory speeds.Paulson et al. [24] worked on development of elasticplastic finite element model for quantification of rolling contact fatigue of bearings made of case carburized steel.It was found that the ratio of the total rolling contact fatigue life of refurbished bearing to that of non-refurbished bearing was more for through hardened bearings as compared to the carburized bearings.Salguero, et al. [25] envisaged tribological issues during dry machining of Aluminum-Copper alloys along with Scanning Electron Microscopy (SEM) & Energy Dispersion Spectroscopy (EDS) analysis to study surface morphology of the material before & after wear.
Krishnaiah, et al. [26] presented industrial application of design of experiment & analysis of data and provided solution to various industrial problems using various techniques such as Taguchi design & analysis, statistical analysis including multiple regression analysis & randomized block design.Mathews [27] worked on design of experiments using MINITAB software.He focused on software approach for solving industrial problems using design of experiments & inferential statistics.Taguchi, et al. [28] worked on quality engineering issues in robust engineering technology used in industry.They focused on industrial application of design of experiments, Taguchi design, orthogonal arrays & signal-tonoise ratio.NSK Bearing Catalogues [29-33] are the vendor catalogues which give detailed information about design dimensions & selection of cylindrical rolling contact bearings used in screw compressors.These catalogues are very important from research point of view to know the existing technology in the field of rolling contact bearings.It also helps the researcher to know the scope of research in a particular rolling contact bearing manufactured by a particular vendor.

Cylindrical Roller Bearings in Screw Compressors
Cylindrical roller bearings (Fig. 1) are widely used in screw compressors in the industry.Screw compressors are used in the industry for compressing process gases & refrigerant gases from suction pressures in the range of 0.01 MPa to discharge pressures in the range of 5 MPa.The suction volume flow rate for the gases flowing in screw compressors ranges from 200 m 3 /h to 100000 m 3 /h.The pressure ratio values used depend on the type of the gas which is compressed.The pressure ratios are in the range of 1.3 to 6.The compression can be carried out in single-stage or multi-stage arrangements.In screw compressors, the bearing loads are generated by gas pressure acting on the spiral rollers.Also, in the process of transmission of torque from one rotor to the other, rotor forces are generated which are transmitted to the bearings.Also, the screw compressors have input gears & timing gears.The radial and tangential forces acting on the gears are transmitted to the bearings.During startup of the screw compressor, loads are generated due to inertia of the spiral rotors.These forces are also transmitted to the bearings.Hence, there is a need to properly design the bearings of the screw compressor considering the criticality of loads & also the space constraints.In screw compressors, a proper combination of cylindrical roller bearings & angular contact bearings is used.The bearings are arranged such that the radial loads are taken only by the cylindrical roller bearings & the axial loads are carried only by the angular contact ball bearings.A correct radial clearance is maintained between the bearing outer raceway and its housing.Due to this radial clearance, the angular contact bearing cannot carry any radial load.Also, the cylindrical roller bearing can carry a very small amount of axial load if subjected to axial loading.
Fig. 1 shows a cylindrical roller bearing used in screw compressors.A low-pressure process gas or refrigerant gas enters the screw compressor through its suction nozzle.It gets compressed by the spiral rollers which rotate continuously at high speeds.The compressed gas discharges through the outlet nozzle of the screw compressor.

Selection of Roller Bearing
In order to decide the dimensions (Table 1) of the cylindrical roller bearing, vendor catalog data of NSK Bearings as shown in the Fig. 2 & Fig. 3 is referred.

Determination of Dynamic Load Carrying Capacity of the Bearing
The dynamic load carrying capacity of a cylindrical roller bearing is the amount of radial load the bearing can carry for a minimum life of one million revolutions.The life of a cylindrical roller bearing is the number of revolutions the bearing will complete just before the first occurrence of a fatigue crack in any of the cylindrical roller or in inner or outer races of the bearing.While calculating the dynamic load carrying capacity of a bearing, it is assumed that the inner race of the bearing is rotating & the outer race is stationery or fixed.The actual dynamic load carrying capacity of the bearing based on the empirical equation mentioned in the NSK bearing vendor catalog is estimated & then compared with the maximum dynamic loading capacity of the bearing mentioned in the same NSK bearing vendor catalog.The bearing life in million revolutions is given by the Eq. 2. (2) The screw compressors used for process gases & refrigerant gases operate at high rotational speeds.The maximum speed of the shaft coupled to the inner race as 3500 RPM is considered.Also, the bearings which are used in machinery for industrial applications require continuous operation of 24 hours per day.In case of pumps & compressors such as screw compressors used for process gas & refrigerant gas compression, the machines run for around 50000 hours.
Also, the radial load acting on the rolling contact bearing is given by Eq.3. ( Substituting the values we get, N Hence, 7179 N is the maximum load that the bearing can carry in the radial direction.

Determination of Loads in Weight Pan for Pin-on-Disk Tribometer
Based on dynamic load carrying capacity of the bearing, the maximum radial load required to be carried by the cylindrical roller bearing is limited to 7179 N.This load will be shared by 12 cylindrical rollers.
= Maximum Normal Load that can be carried by the bearing based on its dynamic load carrying capacity.

= Number of cylindrical rollers per row.
Maximum load that can be carried by each cylindrical roller is: Hence, the maximum normal static load that a cylindrical roller can carry is 2608.37N.This indicates that one can apply maximum 2608.37 N normal load on the roller while conducting simulation on pinon-disk machine considering static & dynamic loading capacity of the bearing.Considering the limitations of the pin-on-disk machine, one can apply maximum load up to 20kg.hence it was decided to select loads as 20 N, 40 N, 60 N, 80 N &100 N.These loads lie within the limiting value of 2608.37 N.

Fig. 4 Determination of disk speeds
The linear speeds of the inner & outer races of the bearing at the point of contact with the roller can be expressed by Eq. ( 5) and Eq. ( 6).
(5) (6) Considering there is no slip at the point of contact of the periphery of the roller & inner & outer races of the bearing (Fig. 4).Therefore, the velocity of the cage & the rolling element is same & can be expressed as the mean value of the linear velocities of the inner & outer cages at the point of contact with the roller.(7) Therefore, the rotational speed of the rollers & the cages about the axis of the shaft can be expressed as: (8) The rotational speed of the roller relative to the inner race can be expressed as: (9) The linear speed of the roller relative to the inner race at the point of contact can be expressed as, 2 60 60 Now equating the linear speed of the roller relative to the inner race at the point of contact with the linear speed of the cylindrical test specimen relative to the rotating disk (Fig. 5) of the pin-on-disk machine at the point of contact. (11) Hence selecting inner race rotating speeds as 1000 RPM, 1200 RPM, 1400 RPM, 1600 RPM & 1800 RPM and consider outer race to be stationary as it is fixed inside the screw compressor.From Eq. ( 14), the speed of rotation of the disk ( ) of the pin-on-disk machine is calculated.Hence, the estimated rotational speeds of the disk corresponding to the inner race rotating speeds from Eq. ( 14) are 692 RPM, 830 RPM, 969 RPM, 1107 RPM & 1246 RPM respectively.

Test Scheme
We adopted "Standard Test Method for Wear Testing with a Pin-on-Disk Apparatus" as mentioned in ASTM G99-95a (Reapproved 2000).

Chemical Composition & Mechanical Properties
EN31 is a commonly used metal in the manufacture of cylindrical rollers & races of antifriction bearings.The purchased EN31 material was tested in a laboratory to determine its chemical composition & mechanical properties (Table 2).

Sample Preparation
25 number of EN31 specimens were manufactured for experimental testing on pin-on-disk machine.The sample size is (15 mm Dia.) x (20 mm length).The specimens were prepared from the raw material by mechanical processes namely machining & heat treatment.The hardness of the specimens at various stages of manufacturing was tested in metallurgical testing laboratory & the values are as presented in Table 3.The EN31 specimens are through hardened.The material of the disc is EN31 & its hardness is 60 HRC.

Lubricant Selection
Lubricant ISO VG32 was selected based on the lubricant selection chart given in NSK bearing chart.

EXPERIMENTATION
Experimentation was performed on pin-on-disk machine to simulate the phenomenon of rolling & sliding contact between the outer surface of inner race of a roller bearing and the cylindrical roller.The hardened EN31 disk of the machine represents the inner race of the bearing & the test specimen held in the fixture which slides against the rotating disk represents the cylindrical roller.As shown in Fig. 6 (a), the main components of the pin-on-disk machine are the rotating disk, friction sensor, wear sensor and linear variable differential transformer (LVDT).The wear, friction force & friction coefficient data generated during experimentation is saved on )  4.

Wear Pattern
During pin-on-disk experimentation, the EN31 steel test specimens experience rolling & sliding motion relative to the rotating disk on the pin-on-disk tribometer.Due to the friction force generated at the interface between the test specimen & the rotating disk, the test specimens get worn out & a specific wear pattern is observed on the peripheral surface of the test specimens.The wear pattern of the samples 1 to 25 used in the experiment number 1 is shown in the Fig. 7.The amount of material removed from a sample due to wear is proportional to the width of the white band appearing on its periphery.From Table 4, it can be seen that at a disk speed of 692 RPM, the amount of material removed increases from    5 in the form of an orthogonal array.
From Fig. 8 (a-e) it is observed that initially the wear rises to a maximum value in a very short time interval & then remains steady for the remaining test duration.From Fig. 9 (a-e) it is seen that initially the friction force rises to a maximum value, falls & then oscillates about a mean value for the remaining test duration.It can also be observed that the friction force changes from positive to negative or vice-versa for some tests.From Fig. 10 (a-e) it is observed that the friction coefficient initially rises to a maximum value in a very short time interval, oscillates & decreases to a minimum value in the remaining test duration.

Statistical Analysis of Experimental Results
S/N Ratio: Signal to Noise ratio measures the variation in response with respect to the control parameters under different noise conditions.
p-value: p-value is a quantitative measure for reporting the result of normality test.A small p-value indicates means that the null hyporhesis is false.
T-value: T-value is the measure of ratio between the coefficient & its standard error.The magnitude of t-value decides whether to reject null hypothesis.Tables 4(a) & 4(b) are the response tables for wear and S/N ratio of wear respectively.These S/N ratios are obtained using smaller-the-better approach.Delta is obtained by calculating the algebraic difference between the maximum & minimum Eqs.(15)(16) are the regression equations for wear & S/N Ratio of wear respectively.From Eqs (15)(16), it is observed that the magnitude of wear as well as S/N Ratio increases with the increase in speed as well as with the increase in load.
In Table 5, T-values are obtained by dividing response coefficients (Coef) by standard error of the coefficients (SE Coef).These T-values are further used to estimate the corresponding p-values.In Table 6, the p-values for all the terms are less than the α value of 0.025 at the significance level of 5%.This indicates that both the predictor variables i.e. speed & load have a significant effect on both the response variables i.e. wear & S/N ratio.The VIF factor is 1 for both the predictors i.e. speed & load which indicates that the predictors are not correlated.
The Table 6  Table 7 shows a result of Analysis of Variance (ANOVA) for wear and S/N Ratio.Higher values of sequential sum of squares & percentage contribution for speed as compared to load indicates that the speed factor accounts for more deviance in the response variable wear as well as in S/N Ratio.From Table 7 it is seen that both the speed & load factors have significant effects on the response variable wear as well as on S/N Ratio.

Fig. 3
Fig. 3 Roller Bearing Dimension Substituting this data in the Eq.(1), = 115455 N From NSK bearing vendor catalog, = 335000 N 115455 N < 335000 N, the design of the roller bearing & the cylindrical rollers is safe from dynamic loading capacity point of view.

Fig. 6 (
Fig. 6 (a) Experimental setup (b) Fixture for holding test specimens (c) Schematic of tribological test configuration 17.410 µm to 118.610 µm as the load varied from 20 N to 100 N. At a disk speed of 830 RPM, the amount of material removed increases from 49.130 µm to 122.650 µm as the load is varied from 20 N to 100 N. At a disk speed of 969 RPM, the amount of material removed increases from 88.200 µm to 118.530 µm as the load is varied from 20 N to 100 N. At a disk speed of 1107 RPM, the amount of material removed increases from 111.700 µm to 144.870 µm as the load is varied from 20 N to 100 N. At a disk speed of 1246 RPM, the amount of material removed increases from 177.060 µm to 196.080 µm as the load is varied from 20 N to 100 N.

Fig. 7 Fig. 8
Fig. 7 Wear pattern of the worn-out test specimens

Fig. 10
Fig. 10 Variation of friction coefficient with time SE coefficient: The standard error of coefficient measures the precision of the estimated coefficient.Smaller the standard error, more precise is the estimated value.AIC & BIC: These are Akaike's Information Criterion & Bayesian Information Criterion respectively.These are the measures of the relative quality of the model & this measure is dependent on number of terms in the model.Smaller values of AIC & BIC are desirable.
shows model summary for regression analysis of wear & S/N Ratio.Higher standard deviation of wear indicates lesser closeness of the observed values to the line of fit.The R 2 values for both wear and S/N Ratio are below 70 % which indicates that the linear fit of the model can be further improved.The difference between & R 2 for both wear & S/N Ratio is below 4% and hence not much significant.It means that there are no insignificant predictors in the model.There is a substantial difference between & R 2 in case of S/N Ratio which indicates that the model is slightly over-fit.The values of AIC & BIC are a bit higher in case of both wear & S/N Ratio and if reduced would give a better linear fit.

Fig. 11
Fig. 11 Pareto charts of standardized effects (a) Response is wear (b) Response is S/N Ratio Fig.11 shows Pareto charts in which the absolute values of standardized effects for the predictors A (speed) and B (Load) are plotted.In Fig.11(a), wear is the response variable.In Fig.11(b), signal-to-noise ratio is the response variable.The bars which cross the reference line are statistically significant.In Fig.11 both the bars cross the reference line which indicates that both the factors speed and load have a significant effect on the response variables wear as well as on signal-to-noise ratio of wear.

Fig. 12
Fig. 12 Half-Normal probability plots (a) Response is wear (b) Response is S/N Ratio

Fig. 13
Fig. 13 Residual versus fitted values (a) Response is wear (b) Response is S/N Ratio

Figs. 14
Figs.14(a-b) show histograms of residuals.In Fig.14(a), the frequency of residuals is higher in the residual ranges (-25 to -35) and (15 to 25) than the frequency at the center.Hence, the histogram deviates from the bell-shaped normal distribution curve.In Fig.14(b), the frequency of residuals is higher in the range of residuals (-2 to -4) than the frequency at the center.Hence, this histogram also deviates from the bell-shaped normal distribution curve.

Figs. 15
Figs.15(a-b) are the interaction plots for data means of the response variables wear & S/N ratio respectively.Load & speed are the predictor variables.These interaction plots show how a response variable varies when one predictor variable is kept constant & the other is varied.In Fig.15(a), the top right plot shows variation in wear with load for five different levels of speed & load.The bottom left plot shows variation in wear with speed for five different levels of load & speed.Similarly in Fig.15(b), the top right plot shows variation in S/N ratio with load for five different levels of speed and load.The bottom left plots show variation in S/N ratio with speed for five different levels of load and speed.

Fig. 15
Fig. 15 Interaction Plots for (a) response is wear (b) response is S/N Ratio S.H. Gawande & Ketan Raykar.Wet Sliding Wear Behavior of EN31 Rolling Contact Bearings for Screw Compressor Application with Statistical Analysis temperature & running time on rolling contact fatigue of bearings.Zhang, et al. [12] solved finite-length sliding contact problem using semi-analytical methods as Conjugate Gradient Method (CGM) and Fast Fourier Transform (FFT).

Table 1
Selection of roller bearing

Table 2
Chemical composition & Mechanical Properties of EN31 material

Table 3
Hardness of EN31 Test specimens

Table 5
Regression Analysis-coefficient table for wear & SN ratio of wear

Table 6
Regression Analysis-model summary for wear & SN ratio of wear

Table 7
Analysis of variance for wear & SN ratio of wear