Recent advances in Predictive Maintenance Techniques for Lubricant Condition Monitoring

1. Introduction

Lubrication is a critical component in maintaining the efficiency and reliability of mechanical systems. The primary function of lubricants is to reduce friction and wear between moving parts, which significantly extends the operational lifespan of equipment. However, over time, lubricant conditions degrade due to contamination, oxidation, and thermal stresses, which can lead to machinery failure if left unchecked. Lubricant condition monitoring (LCM) is thus employed to maintain the health of lubricants to prevent machine malfunctions. Traditional maintenance strategies, such as time-based or reactive approaches, often fail to account for real-time variations in lubricant conditions, leading to unexpected downtime and costly repairs. Predictive maintenance techniques, particularly those involving lubricant condition monitoring, offer a solution by utilizing advanced modeling and sensor data to predict and prevent failures before they occur. This paper details the potential incorporation of different techniques, such as sensors, vibration analytics, and soft computing models, to make LCM more mainstream and easily incorporated into already established systems.

2. Global Performance Passport

LCM refers to proactive maintenance strategies that can assess the current lubricant health or predict the lubricant’s future performance. The monitoring of lubricant properties over time allows for preventative maintenance and helps preserve the lifespan of machinery by preventing equipment failure. Real-time tracking of these properties can reduce maintenance costs by eliminating over-maintenance and the uncertainty that problems with the machinery are a result of lubricant degradation. Thus, it would be beneficial for all industries to adopt tested LCM techniques and share data collected from different applications.
A practical and perhaps traditional approach to standardizing lubricant maintenance and health across industries involves compiling key characteristics and performance metrics into a “global performance passport”4. Just as Safety Data Sheets (SDS) provide workers with essential information about the proper handling and use of potentially hazardous chemicals5, the global performance passport would serve a similar purpose for lubricants. It would contain identifying characteristics of the lubricant, including the lubricant’s name, physical properties and chemical composition. This would allow employees to easily identify the specific type of lubricant in use, especially when working with various machines that require different lubricants. While current lubricant manuals containing this information are typically obtained from the manufacturer, a global performance passport would ideally standardize this information across all industries and lubricant types in a consistent format, making it easier to use, share and apply universally.
Beyond identification, the global performance passport would also include the lubricant’s ideal performance test results and operating conditions as well as maintenance history4. Performance metrics such as viscosity, acidity, and water content would set a standard for the lubricants’ usable range, while operating conditions will define the specific application and environment the lubricant is suitable for. For instance, Figure 1 compares the tribological properties of fresh and worn hydraulic power transmission H46 oil4. The noticeable dip seen in the worn oil’s performance would set a clear threshold for its useability, enabling workers to easily identify when the oil is nearing the end of its effective life. The inclusion of an up-to-date maintenance log would allow for easy tracking of changes over time, so that employees can visually see a decreasing trend in lubricant health and schedule an inspection and replacement in advance.
However, measurements of performance may be too generalized to apply to all types of lubricants, and storing information in a centralized database may leave it vulnerable to data breaches. Different industries and countries may also find it hard to agree on the same set of standards for healthy lubricant use. The system could also be abused through data falsification from companies or individuals attempting to avoid necessary maintenance or repairs. The unification of such information can best be put into place by established standardization organizations. For instance, the American Society for Testing and Materials (ASTM) International is a global organization that develops and delivers voluntary consensus standards with over 12,000 standards in use around the world6. ASTM International already has standards on lubricant properties and testing, but it does not cover everything a global performance passport would provide. If large organizations develop and publish LCM standards, industries that adhere to their policies will follow suit, contributing to the global standardization of LCM.

3. Sensors

Sensors play a critical role in LCM, as they can directly measure key indicators of lubricant health in real-time. This eliminates the need for frequent disassembly of machinery to check for lubricant degradation. Excessive maintenance could cause the loss of lubricant, introduce contaminants to otherwise healthy lubricants, increase downtime, or damage machine components due to constant disassembly and reassembly. Often, lubricants are replaced before the end of their life cycle, but the integration of sensors into LCM would optimize both lubricant and machinery lifespan, reducing waste from unnecessary lubricant changes7.
In railway applications, lubrication is essential for maintaining the smooth operation of moving parts and extending the longevity of equipment. Axle box bearings, in particular, have a high load rating as they support the axles and wheels of high-speed railway vehicles8. Initially, these bearings were lubricated with mineral oil but were later replaced with lithium grease due to heavy oil loss leading to reduced lubricant effectiveness and environmental contamination9.  As shown in Table 1, a commonly used grease in axle box bearings called Arapen RB 320 is made from mineral oil thickened with a lithium calcium soap, resulting in a kinematic viscosity about 5 times greater than that of pure mineral oil at 40°C10,11. The higher viscosity allows for a reduction in initial lubricant amount and loss during operation. LCM for lubricating grease will primarily focus on its water content since water can alter its properties (e.g., reducing viscosity) and cause corrosion in the metal bearings.
Researchers at the Austrian Center of Competence for Tribology tested commercially available humidity sensors in axle box bearings to evaluate their effectiveness and robustness12. These sensors successfully detected increases in humidity–indicating water intake–and withstood challenging railway environmental factors, like thermal cycling, relative humidity, and mechanical load. Figure 2 shows temperature and relative humidity measurement results from random vibration tests and shock tests by comparing the sensor’s measurements to reference values12. The sensor in the random vibration test showed relatively similar outputs with the reference values, while the shock test indicated no change in the sensor’s capabilities when subjected to abrupt acceleration. This suggests that the humidity sensor can perform accurately under stressful conditions, potentially making them a non-invasive tool for LCM in railway applications. While the researchers believe that the sensors will not require additional maintenance beyond regular vehicle maintenance, they have yet to conduct long-term studies on the sensors’ reliability in prolonged stressful conditions. This could be done by increasing the parameters in stressful condition experiments, accelerating the aging process via UV radiation or oxidation, or by conducting virtual simulations. Once possible tests have been undertaken, on-the-field performance results will dictate whether the robustness of the sensors in experiments will carry over to real applications. Furthermore, outdoor operating conditions for railways in varying climates may affect sensor performance, particularly in regions with extreme temperatures or heavy rainfall.
Hydraulic systems, on the other hand, typically use oil as a low-viscosity lubricant to operate at high speeds. Researchers at Dongguk University and Hyundai in Korea developed an integrated oil sensor to monitor hydraulic oil contamination in construction machinery13. The sensor measured absolute viscosity, density, temperature, and the dielectric constant of the lubricant to establish a benchmark for the effects of various pollutants found in construction environments. Figure 3 shows that when small amounts of dust or improper oil were introduced, the lubricant’s absolute viscosity and dielectric constant did not change significantly13. However, the introduction of moisture resulted in measurable variations, particularly in the dielectric constant; more noticeable effects were observed with the presence of varnish–an insoluble deposit formed from oil oxidation14. The dielectric constant reflects the lubricant’s ability to transmit electric currents, which can be affected by contaminants or additives in the lubricant15. Thus, this oil sensor is useful for identifying contaminants that affect the lubricant’s dipole moment, but it is limited in its ability to detect other types of contaminants.

Figure 3: Variations in hydraulic oil properties – absolute viscosity (top) and dielectric constant (bottom)
– when introducing dust (left) and more than 3000ppm of moisture (right)13
 
Although integrated sensors may offer a more efficient method of LCM compared to traditional oil sampling and analysis, long-term testing is still required to assess the sensor’s capability to detect chemical reactions with contaminants over extended periods.

4. Vibration Analysis

An indirect way of LCM is through vibration analysis, which monitors changes in a system’s performance by tracking its established vibration signature16. When a certain vibration signature is associated with normal operation, the deterioration of lubricants essential for machine health will be associated with a change in the vibration signature. Thus, vibration analysis could aid LCM by providing a warning for the presence of degraded lubricants, especially if certain frequencies can be linked to defects in specific locations. The vibration spectrum of a rotating equipment will show peaks at multiples of the fundamental rotational frequency, called harmonics. For instance, a motor running at 1500RPM exhibits synchronous peaks at 1500RPM, 3000RPM, and so on17. Ninety percent of the problems detected by vibration analytics are related to balance and alignment, which causes excessive vibration and eventually wear down bearing surfaces if not addressed18. Thus, unusual variations in amplitude or frequency usually indicate imminent machine failure –identifying these variations allows for repairs before significant malfunctions occur.
Zamorano et al. at the Universidad Carlos III and the Universidad de La Laguna examined the use of vibration analysis on centrifuge lube oil separation systems in marine vehicles19. After preprocessing the vibration data to remove noise, normalizing it, and extracting relevant features, the signals were decomposed into time-frequency components using wavelet packet transform, allowing for the analysis of different frequency bands. The energy levels of the motor were measured twice: the first measurement was taken right after a maintenance task, and the second measurement was taken after several working hours. Figure 5 shows that the motor energy increased with each harmonic in the first measurement but decreased at the second harmonic in the second measurement19. This suggests that there is a possible correlation between changes in vibration patterns and lubricant health, as lubricants degrade with continued system operation. Being able to detect variations in energy levels from vibration data allows for the comparison of patterns and individual values at specific harmonics to reference values, but the relationship between lubricant degradation and change in vibration data must be further studied and confirmed.
Figure 4: Energy variations between first and second measurements of marine oil separation systems19

5. Soft Computing Methods

The widespread use of technology and artificial intelligence (AI) has already significantly improved convenience and accuracy in both personal and business settings. Although still in development, AI can search databases far more quickly than humans, which could be especially useful for new staff with less experience. It also has the potential to perform around-the-clock monitoring if proven reliable against bias and complex situations. Monitoring might function similarly to AI in home security systems, which learns the homeowner’s patterns and sends out alerts if it detects any threats or potential risks. Specifically, soft computing – referring to a branch of AI that relies on learning from data to make “human-like” decisions, focusing on approximate solutions and models while tolerating imprecision, variations. and noise19– presents a promising method for LCM.
Pourramezan et al. at the Ferdowsi University of Mashhad and the Tarbiat Modares University in Iran utilized soft computing methods like K-Nearest Neighbor (KNN) and Radial Basis Function Artificial Neural Networks (RBF-ANN) to predict lubricant condition and engine health using parameters like viscosity, acidity, particle count, and vibration data21. The software was programmed to diagnose engine health conditions in three categories (normal, caution, or critical) based on seven key lubricant indices. After training all models on historical datasets, they found that engine health diagnosis accuracy across all models was at least 97%, with larger training datasets resulting in higher accuracies. Figure 6 depicts the results of the KNN method, which saw the overall lowest accuracy of greater than 96% for lubricant wear when trained on 40% datasets21. Despite the lower accuracy with smaller training sets, soft computing models demonstrated impressive reliability, with potential for further improvement as more data becomes available.
 
Table 2: KNN method for detecting lubricant wear using three sizes of training sets21
In a separate study, the same researchers conducted a comparative analysis of six different soft computing models for predicting the viscosity of diesel engine lubricants22. Viscosity plays a crucial role, especially at high temperatures, where a reduction in viscosity can lead to increased wear on mechanical parts, while an increase often signals oil oxidation and the accumulation of varnish or sludge23.  As shown in Figure 7, an ISO VG46 oil has a viscosity of 100,000 mm2/s at -40°C that is reduced to less than 10 mm2/s at 140°C24. If the viscosity of the lubricant drops too low, high operating temperatures would further decrease it, compromising the lubricant’s effectiveness. The models were trained using a dataset of 555 engine oil analysis reports, which included data on two oil types, metallic and non-metallic elements (indicative of contaminants/additives), and engine operating hours. They found that the radial basis function (RBF) model consistently outperformed the others in terms of accuracy and consistency. Figure 8 shows that as network topology increased, the RBF model displayed a decrease in root mean square error (RMSE) and an increase in efficiency (EF)22. At the max network topology of 35 neurons, the RBF model had an RMSE of 0.20 and an EF of 0.99 during training, and an RMSE of 0.11 and EF of 1 during testing.

Figure 5: Relationship between viscosity of an ISO VG46 oil and temperature24
Figure 6: Impact of network topology on RBF model performance in terms of RMSE and EF during training (left) and testing (right)22
 
By combining soft computing with historical datasets, the software can learn patterns and relationships between parameters, allowing it to identify potential issues early on and send out notifications for preventative maintenance or replacement. While these results paint an optimistic future for AI in LCM, the models still depend heavily on the diversity of situations in their training dataset. Over-reliance on these models could potentially lead to careless mistakes and expose vulnerabilities in AI when faced with unforeseen errors.

6. Conclusion

Real-time LCM through a combination of robust sensors and reliable software will be key to maintaining equipment health and optimizing performance with minimal system disturbances. An essential step towards achieving this is making information about lubricants, particularly those in common use, easily accessible, such as through a global performance passport or shared online databases. While sensors are already commercially available, specific testing must be conducted based on the industry and machinery in which they will be used. Developing a dedicated sensor for every application is impractical and time-consuming, so the ideal approach would be to repurpose existing technologies for new applications. In addition, vibration analysis has to be further evaluated for relevant correlations with lubricant health through studying different machinery across industries.
Although there is ongoing debate about the reliability of AI in LCM, the focus should be on creating software that is flexible enough to adapt to changes yet detailed enough to detect subtle shifts that may indicate current or future lubricant issues. Studies pertaining to the implementation of LCM in current machinery suggest that the process will be non-invasive, eliminating the need for entirely new equipment. Replacing a machine is always more expensive than replacing its lubricant, so incorporating additional systems – though potentially costly upfront – would be more cost-effective than dealing with frequent machine breakdowns caused by missing the critical moments for lubricant replacement.
Even without fully relying on AI programs, it remains an important factor in making LCM incorporation into different industries seamless, a process that should ultimately be guided and overseen by human expertise.

References

1 Lansdown, A. R. (1982). Lubrication: A Practical Guide to Lubricant Selection. http://ci.nii.ac.jp/ncid/BA07131179
2 Lubrication increases the efficiency and life-expectancy of machines. (n.d.). https://www.graco.com/gb/en/vehicle-service/solutions/articles/what-is-lubrication-and-why-is-it-important.html
3 Admin. (2023, December 23). The Six Forms of Lubricant Degradation. Strategic Reliability Solutions Ltd. https://strategicreliabilitysolutions.com/the-six-forms-of-lubricant-degradation/
4 Radulescu, I., & Radulescu, A. V. (2020). Lubricants condition monitoring by using a global performance passport. IOP Conference Series Materials Science and Engineering, 724(1), 012035. https://doi.org/10.1088/1757-899x/724/1/012035
5 Safety Data Sheets. (2022, October 25). Safety Services. https://safetyservices.ucdavis.edu/units/ehs/research/safety-data-sheets#:~:text=The%20purpose%20of%20a%20Safety,regarding%20chemical%20hazards%20and%20handling.
6 Detailed overview - Overview - About Us. (n.d.). https://www.astm.org/about/overview/detailed-overview.html
7 Halme, J., Gorritxategi, E., Bellew, J. (2010). Lubricating Oil Sensors. In: Holmberg, K., Adgar, A., Arnaiz, A., Jantunen, E., Mascolo, J., Mekid, S. (eds) E-maintenance. Springer, London. https://doi.org/10.1007/978-1-84996-205-6_7
8 Axlebox Bearings for Passenger Cars & Locomotives. (n.d.). Schaeffler Group USA Inc. https://www.schaeffler.us/us/products-and-solutions/industrial/industry_solutions/rail/axlebox_bearings_passenger_cars_locomotives/
9 The evolution of railway axlebox technology - Evolution. (2020, January 17). Evolution - Technology Magazine From SKF. https://evolution.skf.com/the-evolution-of-railway-axlebox-technology/
10 Arapen RB 320. (n.d.). https://www.mobil.com/en/lubricants/for-businesses/industrial/lubricants/products/products/arapen-rb-320
11 Palub. (2020, September 1). Understanding the viscosity index of a lubricant. Q8Oils. https://www.q8oils.com/energy/viscosity-index/#:~:text=The%20viscosity%20index%20of%20a%20lubricant%20is%20determined%20by%20measuring,can%20be%20up%20to%20120.
12  Dubek, K., Schneidhofer, C., Dörr, N., & Schmid, U. (2024). Laboratory robustness validation of a humidity sensor system for the condition monitoring of grease-lubricated components for railway applications. Journal of Sensors and Sensor Systems, 13(1), 9–23. https://doi.org/10.5194/jsss-13-9-2024
13 Hong, S. H., & Jeon, H. G. (2022). Monitoring the Conditions of Hydraulic Oil with Integrated Oil Sensors in Construction Equipment. Lubricants, 10(11), 278. https://doi.org/10.3390/lubricants10110278
14 Understanding varnish contamination in rotating equipment – and how to solve it effectively. (n.d.). Pall. https://www.pall.com/en/oil-gas/blog/varnish-in-oil.html
15 Corporation, N. (2012, February 27). How Dielectric Instruments Can Aid Oil Analysis. Machinery Lubrication. https://www.machinerylubrication.com/Read/28778/dielectric-instruments-oil-analysis
16 Dynapar. (2018, March 9). Vibration Analysis and Vibration Monitoring. https://www.dynapar.com/technology/vibration-analysis/
17 Vibration Analysis & Monitoring | Auricle Vibration Services Private Limited. (2019, February 1). Auricle Vibration Services Private Limited. https://www.auricle.in/services/vibration-analysis-and-vibration-monitoring/
18 Fitch, J. (2024, January 23). How Vibration and Oil Analysis compare and contrast | Machinery Lubrication. Machinery Lubrication. https://www.machinerylubrication.com/Read/32579/how-vibration-and-oil-analysis-compare-and-contrast
19 Zamorano, M., Avila, D., Marichal, G. N., & Castejon, C. (2022). Data Preprocessing for Vibration Analysis: Application in Indirect Monitoring of ‘Ship Centrifuge Lube Oil Separation Systems.’ Journal of Marine Science and Engineering, 10(9), 1199. https://doi.org/10.3390/jmse10091199
20 What is soft computing - Javatpoint. (n.d.). www.javatpoint.com. https://www.javatpoint.com/what-is-soft-computing
21 Pourramezan, M. R., Rohani, A., Siavash, N. K., & Zarein, M. (2021). Evaluation of lubricant condition and engine health based on soft computing methods. Neural Computing and Applications, 34(7), 5465–5477. https://doi.org/10.1007/s00521-021-06688-y
22 Pourramezan, M. R., Rohani, A., & Abbaspour-Fard, M. H. (2023). Comparative Analysis of Soft Computing Models for Predicting Viscosity in Diesel Engine Lubricants: An Alternative Approach to Condition Monitoring. ACS Omega. https://doi.org/10.1021/acsomega.3c07780
23 Corporation, N. (2019, January 5). Oil Viscosity - How It’s Measured and Reported. Machinery Lubrication. https://www.machinerylubrication.com/Read/411/oil-viscosity
24 Lubricant | Basic Bearing Knowledge | Koyo Bearings /JTEKT CORPORATION. (n.d.). https://koyo.jtekt.co.jp/en/support/bearing-knowledge/12-2000.html

About the Authors

Dr. Raj Shah is a Director at Koehler Instrument Company in New York, where he has worked for the last 25 plus years. He is an elected Fellow by his peers at IChemE, AOCS, CMI, STLE, AIC, NLGI, INSTMC, Institute of Physics, The Energy Institute and The Royal Society of Chemistry. An ASTM Eagle award recipient, Dr. Shah recently coedited the bestseller, “Fuels and Lubricants handbook”, details of which are available at ASTM’s LongAwaited Fuels and Lubricants Handbook 2nd Edition Now Available https://bit.ly/3u2e6GY.
He earned his doctorate in Chemical Engineering from The Pennsylvania State University and is a Fellow from The Chartered Management Institute, London. Dr. Shah is also a Chartered Scientist with the Science Council, a Chartered Petroleum Engineer with the Energy Institute and a Chartered Engineer with the Engineering council, UK. Dr. Shah was recently granted the honourific of “Eminent engineer” with Tau beta Pi, the largest engineering society in the USA. He is on the Advisory board of directors at Farmingdale university (Mechanical Technology), Auburn Univ (Tribology), SUNY, Farmingdale, (Engineering Management) and State university of NY, Stony Brook ( Chemical engineering/ Material Science and engineering). An Adjunct
Professor at the State University of New York, Stony Brook, in the Department of Material Science and Chemical engineering, Raj also has over 680 publications and has been active in the energy industry for over 3 decades. More information on Raj can be found at https://bit.ly/3QvfaLX
Contact: rshah@koehlerinstrument.com

Dr. Vikram Mittal PhD is an Associate Professor in the Department of Systems Engineering at the United States Military Academy. His research interests include energy modeling, technology forecasting, and engine knock. Previously, he was a senior mechanical engineer at the Charles Stark Draper Laboratory. He holds a PhD in Mechanical Engineering from MIT, an MS in Engineering Sciences from Oxford, and a BS in Aeronautics from Caltech. Dr. Mittal is also a combat veteran and a major in the U.S. Army Reserve.
Ms. Ivy Lu is part of a thriving internship program at Koehler Instrument company in Holtsville and is a student of Chemical and Molecular Engineering at Stony Brook University, Stony Brook, New York where Dr. Shah is on the external advisory board of directors.

Editors

Nicholas Douglas, Daniel Baek, Angelina Mae Precilla, Gavin Thomas