Rolling contact fatigue dataset for AISI 52100 bearing steel balls under calcium grease lubrication.

Background & Summary

Rolling-element bearings are essential parts of rotating machinery, and failure of one or more bearings can lead to expensive downtime and losses. Although bearings can fail in a variety of ways, such as adhesive wear, abrasive wear, fretting, corrosion fatigue, electrically induced pitting, and rolling contact fatigue (RCF), the service life of bearings is designed and standardized around RCF. RCF arises from the application of Hertzian contact stress between ball-raceway surfaces due to repeated cyclic loading. In such conditions, subsurface shearing stresses nucleate cracks at material inhomogeneities or surface asperities, and these cracks grow progressively until surface damage (such as pitting, spalling, or flaking of the material) causes the bearing to fail 1 – 3 . This degradation extends further with increased surface roughness and micropitting, where the first interactions between asperities can trigger the process from mild to severe fatigue 4 . Lundberg and Palmgren 5 developed an analytical model that underlies the probabilistic fatigue and is the standard method that is used worldwide for bearing life prediction. Among bearings, AISI 52100 high-carbon chromium steel balls are the most commonly used bearing material because of their hardness, wear resistance, and fatigue resistance under concentrated contact loading 6 , 7 . Within the industrial sector, rolling contacts are mainly lubricated with greases. Calcium-soap-thickened greases of various NLGI consistency grades are widely used in gearboxes and conveyor systems 8 . The grade determines the rheological characteristics of the grease, such as its ability to form a lubricating film, its lubricant starvation tendency, and consequently bearing fatigue life 9 – 11 .

Currently, there are only a limited number of standardized RCF datasets that have enough repetitions to perform a statistical life estimation. Most published RCF presentations do not provide underlying records of the number of cycles a test was run for or the failure records for the surface image, but rather report aggregate summary statistics. While existing datasets have been primarily leveraged to benchmark fault-detection and prognostic algorithms 12 – 14 , several widely used vibration-based diagnostics datasets exist, including the Case Western Reserve University bearing fault dataset 15 and the NASA IMS bearing prognostics dataset 16 . Existing public RCF datasets generally focus on fatigue life measurements or derived results and often do not provide the combination of raw trial-level fatigue life data, bearing surface imagery, statistical analysis outputs, and detailed per-trial metadata under controlled lubrication conditions. The dataset contains these elements in a structured format and may support reuse in fatigue life analysis, statistical benchmarking, and prognostic model development. Optical evidence of surface damage (such as pitting, cracking, and plastic deformation) is likewise rarely deposited in structured, accessible repository formats, despite its relevance to fault signature development in bearing diagnostics.

The dataset consists of all experimental results from 63 rolling contact fatigue tests on AISI 52100 bearing steel balls that were performed with three calcium-soap-thickened greases of NLGI grades 0, 00, and 000, designated EP0, EP00, and EP000, respectively. Each experiment was performed under the IP 300 standard with a four-ball tribometer at 1470 rpm and under a peak axial load of 5739 N, with the aim of automatically detecting failure by vibration-based termination with a vibration threshold of 0.28 g, where g = 9.81 m/s2. Two-parameter Weibull statistics were used to calculate L₁₀, L₅₀, and L₉₀ fatigue life quantiles for each lubricant, along with other Weibull distribution parameters (such as shape, scale, mean, standard deviation, and interquartile range). In addition to the statistical results, 1,140 high-resolution optical microscopy images were captured following the tests and stored in a hierarchically structured repository organized by lubricant type, trial number, and ball identity, documenting pitting, crack propagation, and plastic deformation across all test conditions.

The per-trial time-to-failure data, Weibull parameters, and surface imagery have different applications for research in tribology, lubrication science, and bearing diagnostics. The NLGI grades differ systematically under the same contact conditions. This provides a controlled basis for studying how the grease’s consistency influences its fatigue life and the damage it experiences. A previous study by Bhaumik and Paleu 17 reported the wear and RCF behaviour of AISI 52100 steel under lubricants with additives using the same test configuration. Another study by Bhaumik et al . 18 offered a broader view of how the bearing material and lubricant behave under alternating electric current. The current dataset expands on this research by providing complete trial data and images of the surfaces after RCF test, which are available as open data for three grease grades. The time-to-failure data from each trial and Weibull parameters can serve as reference data for predicting fatigue life and evaluating fault-detection algorithms. The microscopy images reveal the surface damage for all 63 trials, organized by lubricant grade, trial number, and ball identity. This organization allows for easy retrieval of images under specific test conditions. The data can be accessed on Mendeley Data with the https://doi.org/10.17632/d5n3zgvs7p.3 19 .

Methods

Experimental setup

The four-ball tribometer (Magnum 800-FBT) was used to perform all rolling contact fatigue tests according to the IP 300 standard configuration. In this configuration, one AISI 52100 bearing steel ball is fixed in the upper rotating spindle, and three identical balls of the same grade are positioned in a specially designed frictionless ball pot under the rotating spindle. All tests were performed with AISI 52100 bearing steel balls, 12.7 mm in diameter, and with an HRC hardness of 55 – 60. No balls were used more than once; a new set of four AISI 52100 balls was used for each independent trial.

Ball preparation

Before every test, all balls were ultrasonically cleaned for 10 min in hexane, dried with compressed air, and optically inspected under an optical microscope to ensure the absence of pre-existing surface defects or any scratches. The initial surface roughness of the balls was measured to be Ra = 0.101 µm by a Zeiss Surfcom Touch 50.

Lubricant preparation

In this study, three greases with NLGI consistency grades of 0, 00 and 000 (EP0, EP00, and EP000) were used as lubricants. The basic oil formulation (ISO VG150 cSt) and the extreme-pressure additive package were common to all three greases and the only difference between the three was the concentration of calcium soap thickener, which determines the grade according to the NLGI classification system. This controlled variation eliminates the influence of the type of additive and base oil chemistry on fatigue life and only provides a variation of grease consistency. The ball pot was filled with 35 ± 3 g of grease according to the IP 300 standard specification.

Operating conditions

The test conditions were chosen to be directly relevant to industrial testing conditions, with the upper ball being rotated at a constant spindle speed of 1470 rpm corresponding to the synchronous output speed of a standard four-pole 50 Hz AC induction motor on the same supply. A stepped loading protocol was applied in accordance with the IP 300 standard, with an initial applied load of 981 N for the first ten seconds, allowing the lubricant film to establish across the contact zone, thus preventing adhesive seizure at the start. The load was gradually increased to a peak axial load of 5739 N and maintained at that load for the rest of the test till pitting occurred. The equipment was automatically shut off as soon as the pitting occurred. This four-ball configuration provides high stress cycles per hour at each contact point, allowing for accelerated fatigue life testing under controlled elastohydrodynamic lubrication (EHL) conditions.

Environmental conditions

Testing was done under ambient laboratory conditions. Temperature was actively controlled during the test program and was reported to be 35 ± 3 °C. The relative humidity was not actively controlled but was recorded to be 65 ± 3% across the tests.

Fatigue monitoring and termination criteria

The vibration data at the tribometer spindle housing was continuously recorded during each test with a built-in piezoelectric accelerometer (sensitivity: 98.10 mV/g), which was firmly fixed to the tribometer housing. The amplitude of the analog output of the accelerometer was extracted by the data acquisition system using a Fast Fourier Transform (FFT) and was processed in real time with a sampling frequency of 1000 samples per second. No digital filtering or windowing procedures were applied to the vibration signal beyond the raw FFT computation. The tests were automatically interrupted whenever the monitored amplitude of vibration exceeded the vibration amplitude threshold of 0.28 g, where g = 9.81 m/s². The upper ball was rotating to produce one complete cycle of stress at contact areas of the lower balls; the time-to-failure is reported in seconds. The elapsed time at the point of termination was recorded as the time-to-failure value for that trial. During the test program no run-out suspensions were experienced and all 63 trials were only terminated when the vibration amplitude exceeded the predefined threshold due to the surface being damaged due to pitting.

Microscopic examination and image acquisition

After each test, the three stationary balls were taken out of the ball pot, and the contact surfaces of the balls were analyzed with an Olympus BX53M metallurgical optical microscope. The images were taken using a magnification of 5x in a fixed imaging resolution of 2160 × 1620 pixels, and stored as a.tif image file. The lighting conditions and focus parameters were fixed for all acquisitions, so as to make the images easily comparable. Five to seven images were captured per ball per test, resulting in a total of 1,140 pictures for the 63 tests and three lubricants. Images are stored in the Microscopy_Images folder of the repository, organized hierarchically by lubricant grade (EP-0, EP-00, EP-000), trial number (TEST-1 to TEST-21), and ball identity (BALL-1 to BALL-3).

Number of trials and statistical analysis

In order to obtain a meaningful statistical estimation of fatigue life and to account for the natural scatter of the data in rolling contact fatigue tests, a total of 63 trials were performed: 21 independent tests for each of three grades of calcium soap-thickened grease (EP0, EP00, and EP000). The time-to-failure was taken as the elapsed time in seconds before the vibration-based failure criterion was reached and was plotted against a two-parameter Weibull distribution. This distribution is defined by the cumulative failure probability \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F\left(t\right)$$\end{document} F t at a given time (seconds) t : \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F\left(t\right)=1-\exp \left[-{\left(\frac{t}{\eta }\right)}^{\beta }\right]$$\end{document} F t = 1 − exp − t η β where β β is the Weibull shape parameter (slope), representing the failure mode, and η η is the characteristic life (scale parameter) at which 63.2% of the specimens are expected to fail.

The failure data for each lubricant grade were ranked in ascending order, and the cumulative failure probability was estimated using median rank regression via Benard’s approximation: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\rm{Median\; Rank}}\approx \frac{i-0.3}{n+0.4}$$\end{document} Median Rank ≈ i − 0.3 n + 0.4 where i i is the rank of the failure and n n is the total number of trials per grade ( n=21 n = 21 ). The Weibull parameters were derived using Minitab by linearizing the cumulative distribution function into the linear form \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$y={mx}+c$$\end{document} y = mx + c , where the slope of the regression line corresponds to the shape parameter β β .

Using the established shape and scale parameters, specific fatigue life quantiles \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${L}_{10}$$\end{document} L 10 , \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${L}_{50}$$\end{document} L 50 , and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${L}_{90}$$\end{document} L 90 representing the life percentile at which 10%, 50%, and 90% of specimens fail were calculated analytically using the inverse cumulative distribution function: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${L}_{p}=\eta {\left(-\mathrm{ln}\left(1-\frac{p}{100}\right)\right)}^{1/\beta }$$\end{document} L p = η − ln 1 − p 100 1 / β where p p is the percentile of interest (it indicates the percentage of specimens (bearing steel balls) that are expected to fail at or before the calculated cycle count). This methodology allowed for a comparative analysis of how calcium grease consistency influences the fatigue life and damage progression of AISI 52100 bearing steel under high-pressure Hertzian contact. All 21 trials per lubricant grade were included in the analysis; no tests were excluded due to unusual behavior or equipment faults. The overall experimental workflow, from lubricant preparation to Weibull analysis, is summarized in Fig. 1 . Fig. 1 Experimental workflow for rolling contact fatigue testing of AISI 52100 bearing steel balls. The flowchart depicts six sequential stages: (1) four-ball tribometer configuration per IP 300, with the upper ball rotating against three stationary lower balls; (2) stepped load application from 981 N to 5739 N at 1470 rpm; (3) progressive surface deterioration through pitting, plastic deformation, and crack formation; (4) vibration detection by an accelerometer mounted on the tribometer housing; (5) automated test termination when the monitored vibration amplitude exceeds the termination threshold (0.28 g, where g = 9.81 m/s2), and recording of duration to failure; and (6) post-test optical microscopy of ball contact surfaces and Weibull-based fatigue life calculation.

Tools used

Tool / instrument Purpose Four-Ball Tribometer Fatigue loading and test execution Accelerometer Real-time vibration monitoring Optical microscope Post-test surface examination Minitab (Least Squares Estimation) Two-parameter Weibull distribution analysis and fatigue life quantile derivation

Data Record

The complete dataset is stored in Mendeley Data under https://doi.org/10.17632/d5n3zgvs7p.3 19 . The dataset consists of four file types: plain text (.txt, Readme file), lossless raster images (.tif, microscopy images), per-trial meta data (RCF_Trial_Data.docx) and portable network graphics (.png, Weibull analysis plots). The root folder is named as Four_Ball_Tribometer_Dataset, which contains an Analysis_graph folder, a Microscopy_Images folder, and a Readme.txt file which explains the contents of the repository. The dataset describes the RCF performance of AISI 52100 bearing steel balls tested under three calcium-soap-thickened greases of NLGI grades 0, 00, and 000, designated as EP0, EP00, and EP000, respectively. The dataset includes processed fatigue life calculations, Weibull distribution plots, and microscopic images acquired after four-ball tribometer tests for documenting surface-initiated damage. All experiments were conducted in accordance with the IP 300 standard. An overview of the dataset and repository is provided in Table 1 . Table 1 Overview of the dataset and repository. Item Description Subject Mechanical Engineering; Materials Science Specific subject area Rolling contact fatigue, bearing steel, tribological testing, fatigue life estimation Type of data Experimental data; optical microscopy images (.tif); Weibull distribution plots (.png); per-trial metadata record (.docx) Data collection Four-ball tribometer (IP 300 standard), using AISI 52100 steel balls (12.7 mm diameter) with axial load incrementally increased from 981 N to 5739 N at 1470 rpm. Vibration-based automated termination, post-test optical microscopy at 5 × magnification. L₁₀, L₅₀, and L₉₀ derived by two-parameter Weibull analysis Repository Mendeley Data, https://doi.org/10.17632/d5n3zgvs7p.3 Related research article Bhaumik, S. & Paleu, V. Wear and rolling contact fatigue analysis of AISI 52100 bearing steel in presence of additivated lubricants. Metals 11, 907 (2021) 17 . The difference is that the above paper investigates mineral oil rather than grease.

The main repository folders and their contents are summarized in Table 2 . Table 2 Main repository folders and contents. Folder Name Description Analysis_graph Distribution overview plots with Weibull parameters (shape, scale, mean, standard deviation, IQR (interquartile range)) and four subplots (probability density function, Weibull probability plot, survival function, hazard function) for each lubricant grade Microscopy_Images High-resolution optical microscopy images (.tif) organized in subfolders by lubricant type, trial number, and ball identity RCF_Trial_Data.docx Complete per-trial metadata record containing time-to-failure values (seconds and cycles) for all 63 trials across the three lubricant grades, with associated test conditions Readme.txt Detailed description of the dataset, file contents, and experimental parameters

The files available in the Analysis_graph folder are listed in Table 3 . Table 3 Files available in the Analysis_graph folder. File Name Description RCF_EP0_Weibull_Statistics.png Distribution plot with four subplots (PDF, Weibull probability, survival, hazard) for EP0 lubrication, together with a statistics table of Weibull parameters (shape, scale, mean, standard deviation, IQR) RCF_EP00_Weibull_Statistics.png Distribution plot with four subplots for EP00 lubrication, together with a statistics table of Weibull parameters RCF_EP000_Weibull_Statistics.png Distribution plot with four subplots for EP000 lubrication, together with a statistics table of Weibull parameters

Each trial yields a single time-to-failure value, as the vibration termination criterion applies to the test as a whole; all three lower balls in a given trial are therefore associated with the same elapsed time. Time-to-failure values are reported in seconds throughout the test program.

The temperature was controlled during the test and was set to be 35 ± 3 °C. A representative subset of the per-trial metadata is provided in Table 4 ; the complete 63-row metadata record is available as RCF_Trial_Data.docx, deposited in the Mendeley Data repository ( https://doi.org/10.17632/d5n3zgvs7p.3 ). A summary of the Weibull statistical outputs for all three lubricant grades is provided in Table 5 . Table 4 Representative per-trial metadata (one trial per lubricant grade shown; full record in RCF_Trial_Data.docx). Trial No. EP0 (NLGI Grade 0) EP00 (NLGI Grade 00) EP000 (NLGI Grade 000) Failure Criterion Temperature (°C) Time to Failure (s) Cycles to Failure Time to Failure (s) Cycles to Failure Time to Failure (s) Cycles to Failure Vibration Threshold g = 9.81 m/s² T-01 2945 72,153 7352 180,124 8822 216,139 0.28 g 35 ± 3 T-02 3121 76,465 3395 83,178 6495 159,128 0.28 g 35 ± 3 T-03 6593 161,529 3428 83,986 7941 194,555 0.28 g 35 ± 3 Table 5 Summary of two-parameter Weibull statistical outputs for each lubricant grade. Parameter EP0 (NLGI 0) EP00 (NLGI 00) EP000 (NLGI 000) Weibull Shape (β) 3.19906 4.76157 2.48492 Weibull Scale η (s) 4359.86 6203.64 5580.01 L₁₀ 2157 3860 2255 L₅₀ 3887 5744 4814 L₉₀ 5658 7398 7805 Mean (s) 3904.87 5679.88 4950.22 StDev (s) 1339.73 1360.30 2129.74

Time-to-failure values are reported in seconds.

Microscopy image structure

This folder contains all raw visual data and is hierarchically organized according to lubricant type, trial number, and ball number. Each subfolder is an identifier to a unique fatigue test, where the parent folder name indicates the lubricant grade (EP-0, EP-00, or EP-000). The TEST-N subfolder denotes the trial number (1–21), and BALL-1, BALL-2, and BALL-3 are used to identify which of the three stationary lower balls was imaged during post-test examination. To standardize the dataset structure, all image files utilize the structural naming template I-[ImageNumber] 5X.tif. Below is the entire hierarchy of the data listed in the repository.

All microscopy images were acquired at 50× magnification (5× objective lens) with a fixed resolution of 2160 × 1620 pixels in.tif format. These images were captured using an Olympus BX53M metallurgical optical microscope. Lighting conditions and focal parameters were held constant across all 1,140 acquisitions to ensure inter-image comparability. Full details of the image acquisition setup are provided in the Methods section.

Technical Validation

The reliability of the dataset is addressed across six dimensions: instrument calibration, experimental repeatability, Weibull goodness-of-fit, outlier handling, microscopy image quality, and comparison with published data.

Instrument calibration

The Magnum 800-FBT four-ball tribometer was calibrated before the test program. This was carried out according to the manufacturer’s four-ball calibration schedule. The load cell and rotational speed controller were checked against the valid and current calibration certificates provided by Magnum Engineers (Normal Load: Sensotronics Beam Type Load Cell, Model 65023-0ITI-NC-00X, Serial No. 89777379; Frictional Torque: IPA Binocular Type Load Cell, Model BR021H2, Serial No. 166422; Speed: Digital Tachometer, range 1–5000 rpm). During the test program, applied loads and spindle speeds were verified within the accuracy tolerances of ± 2% of measured normal load and ± 0.05% FSD (Full Scale Deflection) for speed throughout the test program.

Repeatability of fatigue life measurements

Because rolling contact fatigue (RCF) is stochastic, there is inherent scatter in cycles-to-failure data. The coefficient of variation (CV = [standard deviation / mean] × 100%) provides a quantitative measure of this scatter, allowing for a comparison of fatigue life variability across different lubricant grades. The mean and standard deviation for 21 trials of EP0 were 3904.87 seconds and 1339.73 seconds, respectively, giving a CV of 34.309%. The mean value of EP000 was 4950.22 seconds, while the standard deviation was 2129.74 seconds, resulting in a CV of 43.02%. For EP00, the mean value was 5679.88 seconds and the standard deviation was 1360.30 seconds, yielding a coefficient of variation (CV) of 23.949%. The observed CV values are consistent with the variability typically reported in RCF studies of AISI 52100 bearing steel, with EP00 exhibiting comparatively lower dispersion (23.95%). Twenty-one independent tests per lubricant grade are consistent with the upper range of sample sizes reported in four-ball RCF element studies and are sufficient for stable two-parameter Weibull parameter estimation 20 .

Weibull goodness-of-fit

The two-parameter Weibull distribution was used to fit the cycles to failure data for each lubricant grade; the Anderson-Darling (AD*) statistic and Weibull probability plot correlation coefficient were used to assess the quality of the fit. The two-parameter Weibull distribution was fitted to all three lubricant conditions. For EP0, the fitted parameters were β = 3.199 and η = 4,359.86 seconds (AD* = 4.532, r = 0.883), while EP000 yielded β = 2.485 and η = 5,580.01 seconds (AD* = 2.260, r = 0.912), and EP00 produced β = 4.762 and η = 6,203.64 seconds (AD* = 1.222, r = 0.958). Among the three, EP00 demonstrated the best Weibull fit, as evidenced by the lowest AD* value (1.222) and highest correlation coefficient (r = 0.958), followed by EP000 (AD* = 2.260, r = 0.912) and EP0 (AD* = 4.532, r = 0.883). The increasing β values across EP000, EP0, and EP00 suggest progressively steeper failure rate distributions, while the higher η values for EP00 and EP000 indicate longer characteristic fatigue lives compared to EP0. The two-parameter Weibull analysis, including estimation of the L₁₀, L₅₀, and L₉₀ percentiles, was performed in Minitab 18 using the LSXY (Least Squares in X and Y) estimation method. The analysis reported point estimates of the Weibull parameters and fatigue-life percentiles. The Anderson–Darling (AD) statistic and Weibull probability plot correlation coefficient (r) were used as the primary indicators of goodness-of-fit for each lubricant grade. While the fitted Weibull shape parameters (β > 1) for all three lubricant grades are higher than the traditional Lundberg-Palmgren baseline for bearings (β ≈ 1.1), they reflect the broader variability of dispersion exponents noted in literature; specifically, Tallian 21 highlights evidence that β can increase in rolling contacts due to varying applied loads and the presence of a fatigue-limit effect.

Outlier and anomalous test handling

A total of 21 independent tests were carried out for each lubricant grade, and these were all included in the Weibull analysis. No trials were eliminated due to abnormal vibration, early termination or equipment malfunction. There were no run-out suspensions and all trials made it to the threshold of vibration termination, indicating the loading and vibration conditions applied were adequate to cause fatigue failure within the test program.

Microscopy image quality

The lead author visually reviewed all 1,140 post-test optical microscopy images during image acquisition to ensure adequate focus, absence of motion blur, and sufficient contrast for reliable identification of damage features. All images were accepted without rejection and there were no reacquisitions due to poor minimum image acquisition quality. The uniformity of the magnification used (50×), resolution (2160 × 1620 pixels) and the level of light used on all images makes it possible to compare images across different lubricant grades and trial numbers.

Comparison with published data

The characteristic fatigue life (4359.86 seconds for EP0, 6203.64 seconds for EP00, and 5580.01 seconds for EP000) determined in this study is consistent with the characteristic fatigue life of AISI 52100 lubricated with base mineral oil under the same four-ball tribometer configuration and load schedule as reported by Bhaumik and Paleu 17 . Bhaumik and Paleu used oil lubrication instead of grease, but the order of magnitude agreement of the two studies under the same contact geometry and loading conditions is mentioned to support the credibility of the experimental methodology and the validity of the deposited data. The observed improvement in characteristic fatigue life when the EP0 grease was replaced by the EP000 grease is physically consistent with the low consistency grade of EP000, which helps to improve the flow of the lubricant and replenishment of the lubricant film at the contact zone under the operating conditions applied 9 , 10 .

Usage Notes

This dataset contains 63 independent controlled laboratory rolling contact fatigue tests conducted on AISI 52100 bearing steel balls with three calcium-soap-thickened greases (NLGI grades 0, 00, and 000). The user should be aware that the dataset is applicable to this bearing steel grade, the IP 300 four-ball contact geometry and the stated load escalation profile. Extrapolation to other contact geometry, materials, and loading regimes should be done with proper care.

The microscopy images are organized hierarchically within the Microscopy_Images folder. The experimental variables are presented in the folder names: lubricant grade (EP-0, EP-00, EP-000), trial number (TEST-1 to TEST-21) and ball identifier (BALL-1 to BALL-3). The folder names follow the dataset naming convention (EP-0, EP-00, EP-000), corresponding to the lubricant grades EP0, EP00, and EP000 referred to throughout this paper. The Analysis_graph folder includes the Weibull distribution plots and tabulated statistical parameters for each lubricant grade. These parameters can be directly used to extract L₁₀, L₅₀, and L₉₀ values, or they can be calculated independently from the analytical expression \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${L}_{p}=\eta {\left(-\mathrm{ln}\left(1-\frac{p}{100}\right)\right)}^{1/\beta }$$\end{document} L p = η − ln 1 − p 100 1 / β with the reported shape and scale parameters. To enable those users who want to retrieve data for specific trial or under specific lubricant conditions, the complete per-trial metadata record is provided as RCF_Trial_Data.docx (deposited in the Mendeley Data repository, https://doi.org/10.17632/d5n3zgvs7p.3 ), with each trial’s time-to-failure value corresponding to its associated image folder paths.

The data presented here were collected under a regulated temperature of 35 ± 3 °C. Relative humidity was not actively controlled but was recorded at 65 ± 3% across the tests; users should account for this when comparing results obtained under controlled humidity conditions. Future extensions of this dataset could include a larger number of trials per grade, varying load profiles, or alternative bearing steel grades, all of which would broaden its application.

Ethics statement

This work does not involve any human subjects, animal experiments, or data collected from social media platforms.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Acknowledgements

The authors are grateful to Mr. Shail Mavani and Mr. Sunil Mehta from M/s Mosil Lubricants Pvt. Ltd., Mumbai, India, for providing the grease samples used in this work.

Author contributions

Shubrajit Bhaumik: Data curation, conceptualization, methodology, supervision, validation, writing, review and editing. Prasanna Kumar R: Supervision, validation, writing, review and editing. Rahul Bharathwaj: Conceptualization, methodology, data curation, writing and editing.

Funding

The authors gratefully acknowledge Sri Mata Amritanandamayi Devi (Amma), Chancellor, Amrita Vishwa Vidyapeetham, for her inspiration and for providing financial support for the Article Processing Charges (APC) of this publication. Open access funding provided by Amrita Vishwa Vidyapeetham.

Data availability

The dataset is publicly available in Mendeley Data under https://doi.org/10.17632/d5n3zgvs7p.3 . It can be accessed directly via the provided URL without any special permissions. The repository contains the root folder Four_Ball_Tribometer_Dataset with the Analysis_graph folder, the Microscopy_Images folder, Readme.txt file and the RCF_Trial_Data.docx file containing the complete per-trial metadata record.

Code availability

No custom code was developed or deposited as part of this dataset. The Weibull statistical analysis was performed using Minitab. The resulting distribution plots and tabulated parameters are deposited directly in the Analysis_graph folder of the repository.

Competing interests

The authors declare no competing interests.

References

1. Sadeghi F Jalalahmadi B Slack TS Raje N Arakere NK A review of rolling contact fatigue J. Tribol. 2009 131 041403 10.1115/1.3209132 Sadeghi, F., Jalalahmadi, B., Slack, T. S., Raje, N. & Arakere, N. K. A review of rolling contact fatigue. J. Tribol. 131 , 041403, 10.1115/1.3209132 (2009).

2. Morales-Espejel GE Gabelli A de Vries AJC A model for rolling bearing life with surface and subsurface survival—tribological effects Tribol. Trans. 2015 58 894 906 10.1080/10402004.2015.1025932 Morales-Espejel, G. E., Gabelli, A. & de Vries, A. J. C. A model for rolling bearing life with surface and subsurface survival—tribological effects. Tribol. Trans. 58 , 894–906, 10.1080/10402004.2015.1025932 (2015).

3. Fernandez Rico JE Hernandez Battez A Garcia Cuervo D Rolling contact fatigue in lubricated contacts Tribol. Int. 2003 36 35 40 10.1016/S0301-679X(02)00097-X Fernandez Rico, J. E., Hernandez Battez, A. & Garcia Cuervo, D. Rolling contact fatigue in lubricated contacts. Tribol. Int. 36 , 35–40, 10.1016/S0301-679X(02)00097-X (2003).

4. Mosleh M Bradshaw KK Smith ST Belk JH Shirvani KA Roughness effect in micropitting and rolling contact fatigue of silicon nitride Ceramics 2019 2 135 147 10.3390/ceramics2010013 Mosleh, M., Bradshaw, K. K., Smith, S. T., Belk, J. H. & Shirvani, K. A. Roughness effect in micropitting and rolling contact fatigue of silicon nitride. Ceramics 2 , 135–147, 10.3390/ceramics2010013 (2019).

5. Lundberg G Palmgren A Dynamic capacity of rolling bearings Acta Polytechnica, Mech. Eng. Ser. 1947 1 3 1 50 Lundberg, G. & Palmgren, A. Dynamic capacity of rolling bearings. Acta Polytechnica, Mech. Eng. Ser. 1 (3), 1–50 (1947).

6. Bhadeshia HKDH Steels for bearings Prog. Mater. Sci. 2012 57 268 435 10.1016/j.pmatsci.2011.06.002 Bhadeshia, H. K. D. H. Steels for bearings. Prog. Mater. Sci. 57 , 268–435, 10.1016/j.pmatsci.2011.06.002 (2012).

7. Harris, T. A. & Kotzalas, M. N. Rolling Bearing Analysis: Essential Concepts of Bearing Technology , 5th edn (CRC Press, Boca Raton, 2006).

8. Lugt, P. M. Grease Lubrication in Rolling Bearings (Wiley, Chichester, 2012).

9. Lugt PM A review on grease lubrication in rolling bearings Tribol. Trans. 2009 52 470 480 10.1080/10402000802687940 Lugt, P. M. A review on grease lubrication in rolling bearings. Tribol. Trans. 52 , 470–480, 10.1080/10402000802687940 (2009).

10. Cousseau T Graça B Campos A Seabra J Friction torque in grease lubricated thrust ball bearings Tribol. Int. 2011 44 523 531 10.1016/j.triboint.2010.06.013 Cousseau, T., Graça, B., Campos, A. & Seabra, J. Friction torque in grease lubricated thrust ball bearings. Tribol. Int. 44 , 523–531, 10.1016/j.triboint.2010.06.013 (2011).

11. Cann PM Starvation and reflow in a grease-lubricated elastohydrodynamic contact Tribol. Trans. 1996 39 698 704 10.1080/10402009608983569 Cann, P. M. Starvation and reflow in a grease-lubricated elastohydrodynamic contact. Tribol. Trans. 39 , 698–704, 10.1080/10402009608983569 (1996).

12. Smith WA Randall RB Rolling element bearing diagnostics using the Case Western Reserve University data: a benchmark study Mech. Syst. Signal Process. 2015 64 100 131 10.1016/j.ymssp.2015.04.021 Smith, W. A. & Randall, R. B. Rolling element bearing diagnostics using the Case Western Reserve University data: a benchmark study. Mech. Syst. Signal Process. 64 , 100–131, 10.1016/j.ymssp.2015.04.021 (2015).

13. Lei Y Lin J He Z Zuo MJ Condition monitoring and fault diagnosis of planetary gearboxes: a review Measurement 2014 48 292 305 10.1016/j.measurement.2013.11.012 Lei, Y., Lin, J., He, Z. & Zuo, M. J. Condition monitoring and fault diagnosis of planetary gearboxes: a review. Measurement 48 , 292–305, 10.1016/j.measurement.2013.11.012 (2014).

14. Akhil, V. M., Aravind, S. L. & Nayak, R. Comparison of fault detection data from defective ball bearings using artificial neural networks. in Advances in Manufacturing, Automation, Design and Energy Technologies (eds Sivaram, N. M., Sankaranarayanasamy, K. & Davim, J. P.) 589–598, 10.1007/978-981-99-1288-9_61 (Springer, Singapore, 2023).

15. Loparo, K. A. Bearing Data Center. Case Western Reserve University https://engineering.case.edu/bearingdatacenter (2024).

16. Lee, J., Qiu, H., Yu, G., Lin, J. & Rexnord Technical Services. Bearing Data Set. IMS, University of Cincinnati. NASA Prognostics Data Repository, NASA Ames Research Center, Moffett Field, CA. https://phm-datasets.s3.amazonaws.com/NASA/4.+Bearings.zip (2007).

17. Bhaumik S Paleu V Wear and rolling contact fatigue analysis of AISI 52100 bearing steel in presence of additivated lubricants Metals 2021 11 907 10.3390/met11060907 Bhaumik, S. & Paleu, V. Wear and rolling contact fatigue analysis of AISI 52100 bearing steel in presence of additivated lubricants. Metals 11 , 907, 10.3390/met11060907 (2021).

18. Bhaumik S Exploring the failures of deep groove ball bearings under alternating electric current in the presence of commercial lithium grease Technologies 2025 13 275 10.3390/technologies13070275 Bhaumik, S. et al . Exploring the failures of deep groove ball bearings under alternating electric current in the presence of commercial lithium grease. Technologies 13 , 275, 10.3390/technologies13070275 (2025).

19. Bhaumik, S., Kumar, P. & Bharathwaj, R. Experimental dataset on rolling contact fatigue life of bearing steel under semi-fluid grease lubrication conditions EP0 EP00 EP000. Mendeley Data V3 , 10.17632/d5n3zgvs7p.3 (2026).

20. Johnson, L. G. The Statistical Treatment of Fatigue Experiments (Elsevier, Amsterdam, 1964).

21. Tallian TE Simplified contact fatigue life prediction model — Part I: review of published models J. Tribol. 1992 114 207 213 10.1115/1.2920875 Tallian, T. E. Simplified contact fatigue life prediction model — Part I: review of published models. J. Tribol. 114 , 207–213, 10.1115/1.2920875 (1992).