Executive Summary
Most TLT readers agree that artificial intelligence (AI) has the potential to greatly impact the oil analysis field; some point out that the current technology has some limitations. Many readers consider it to be a helpful part of their oil analysis toolbox, used for harnessing large datasets, analyzing trends quickly and flagging unexpected results before failure. Having a human expert is still essential for interpreting results and exercising oversight. Overall, AI is likely to continue evolving in ways that make it a more effective tool for oil analysis and related applications.

Q.1. How can the role of AI impact the trending of oil analysis results?

Predictive analysis underpinned by sensor-based technology with Internet of Things (IoT) platforms can be a critical enabler for protecting assets while reducing downtime and optimizing efficiency, and enhancing oil life.

Not sure if it can. It will be interesting to see how AI can interpret results and trend analysis given the varying base oil technologies being used in hydraulic fluids and gear oils.
 
Better used oil analysis results.

It should help the results be more accurate on trending analyses, etc.

It should be able to provide an overview, but that would require three-plus rounds of testing.

Only as the results follow data collection. Over the longer term, the technical knowledge gleaned from real world experience will be lost to human sources. Analysts will tend to become lazy as rational analytical processes will be deemed unnecessary. Have we learned nothing with the demise of customer service and relationships?

I think AI will result in a more easily compiled result set for trending.

Better trend analysis is a likely use for this technology, though without very clear rule sets, I can foresee a lot of unnecessary work generated by misinterpreted data. We’re still seeing “AI” telling people to eat glue and swap out table salt for sodium bromide, so blindly trusting it is a recipe for disaster.

It can provide predictions of trouble or failure faster.

AI, or rather machine learning (ML), will have an impact on oil analysis results in two ways. The first is that trending and data extraction will become much faster through AI/ML automation. Second, AI/ML will rely on historical data, which should improve the predictability of future lubricant performance from analysis results.

AI should be taught to review oil analysis as historical learnings have taught us. AI could review a report and indicate areas of concern. Then AI could trend reports for a single unit providing predictive maintenance. Where I think AI could really benefit oil analysis is within fleets with many reports trending. It should be able to quickly tell a maintenance manager how the fleet is doing, trends to watch, etc., all with the click of a button in a quick manner.

Both in a positive and a negative way. Positive: trending becomes easier, especially for not well-trained staff. Negative: wrong results/false correlations will not be encountered anymore that easily (especially by inexperienced staff).

AI may be able to correct human errors during preparation of oil analysis results.

AI could be a good tool to a.) watch the trends and flag “that” result, and b.) well trained AI could evaluate data and summarize results.

ML to deliver faster (and earlier) interpolation of data points, and predictions.

AI has the potential to improve the accuracy of oil analysis trending. It also has the potential to bring attention to trends we had not previously identified. On the flip side, if AI data is not validated, then there is always the potential for contamination of the data set we are making decisions with. The human factor remains with inspections and feedback on the validity of the AI predictions.

I think AI will just take the place of algorithms that have been manually implemented to predict oil trends.

It can help flag the potential issues due to “seeing” it faster.

For a newly constructed system or a forward-thinking organization that is willing to implement the additional equipment necessary for AI to be useful, AI can significantly impact the trending of oil analysis results. For older systems where a retrofit is required for the addition of the equipment, and significant expense and downtime are required for the retrofit, it is not as useful.

I would say that oil analysis already employs a significant number of algorithms built by humans, and AI can only marginally speed up their usage, which is already one of the faster steps in processing oil samples. For AI to develop other algorithms to help speed interpretation of data, I believe it requires the kind of data that does not accompany the oil samples, namely what the end-user is doing or not doing differently to the assets from one sample to the next.

Analysis by a computer instead of a person.

AI will increase the usefulness of oil analysis, making it a more solid and frequently used tool in research and design (R&D) work.

AI could be used to analyze trends in equipment and shape the way that different maintenance regimens are applied.

Provide more timely and accurate analysis. Eliminate current sampling methods via sensor analysis.

1.) Prediction of trending and early indication of failure of systems and components, by comparing the available data. 2) Correlates various trending data and user experience, and brings the real facts to make a decision by root cause analysis (RCA) and failure analysis specialists. 3.) Providing new vital methods and standards to the unknown engineers to understand much better comparison of trending, interpretation and analysis of data and take a proactive decision.

Companies will rely on AI to give them answers instead of having the knowledge or experience base to make informed decisions.

Interpreting oil analysis results properly involves an incredible complex rules engine. AI will allow interpretations to be even more perspective and consider data points beyond the laboratory results.

AI may use a statistical method to summarize data. I expect the response to be pale. AI should not be used for documentation.

It can provide more insight if it has truly learned all the details about the tests and duty service.

Optimizing lubricant formulation, base stock and additives.

The impact is not yet significant.

It speeds up the extraction of meaningful data.

Classification of analysis data, such as oil system debris, becomes easier and efficient. This enables us to analyze more and more trend data of operating machines and provides us more insights.

Faster and more accurate analysis from trends seen in the results

It may be possible that AI will be able to test oil samples more accurately than the way that they are being tested now.

AI can make use of historic data to identify trends. Alerts can be created and sent out while equipment is in operation. Maintenance teams can make follow-ups with minimum loss of production.

If there is an inconsistent approach, then a remedy can be expected in advance.

The AI model will need to be pre-trained on the interpretation of wear metals detected in oil analysis. High zinc levels will indicate imminent failure of bearings. Companies who have required data for training the model treat it as highly confidential information, so it needs a secure data depository for handling such information. After model has been trained on a few thousand cases, it can be confidently used to test failures which is of great help in maintenance of machinery.

AI can interpret complex, nonlinear patterns. Instead of one-size-fits-all alarm limits, AI can calculate equipment-specific baselines and dynamic limits.

High technology ML sensors using AI can help monitor conditions of oils in service if not advanced tests but basic test, which can give an overview of oil condition when it’s in live zone. This can help for proactive maintenance planning and avoid reactive maintenance.

AI enhances oil analysis trending by filtering out lab noise, detecting subtle patterns and forecasting equipment health with predictive models. Instead of static alarm limits, it applies adaptive baselines and correlates oil data with other condition monitoring inputs to deliver faster, more accurate insights. This shift turns oil analysis from a reactive report into a proactive tool for optimizing maintenance, extending oil life and preventing costly failures.

It can help eliminate normal samples so people looking at the rest have more time to look at them.

With online measurement, by gathering and analyzing the data.

AI can give perfect analysis and good monitoring of by 1.) predicting current state of a lubricating oil in use, and 2.) giving the exact lifespan of the oil.

Once a proper database is available, the AI system may recombine results of various tests to extract new information that may help to foresee the future development of lubricant degradation.

Maybe for better and more easy understanding of measurement precision and accuracy for single laboratories as well as cross-laboratory comparison, as well as for interpretation on the cross-trends of multiple parameter changes and their correlated effect to the system at hand.

Will AI change how oil analysis results are interpreted?
Yes 72%
No 28%
Based on an informal poll sent to 15,000 TLT readers.

A great complement, as AI quickly can learn to recognize green/OK/on specification data, freeing up time for the human operators to work on the tricky cases. Especially useful for trending and analytics. Remember, AI is already here, and your competitors might have implemented it even if you have not.

The possibility to use a larger amount of data will open up new possibilities. Instead of just looking at one value, you can look on the full spectrum or chromatograms to get indication of degradation. To get fast analysis of big dataset will improve the value of oil analysis.

AI will facilitate the sample scheduling and data collection and will also be able to analyze the oil analysis data trends and arrive at a result while the analysis is ongoing. It will free the researcher or manager to conduct time consuming analysis. Of course, the ability of the AI depends on the sophistication of the AI software. Some can develop test reports automatically, or if the data do not fit the trend, repeat analysis could be performed in real time.

AI should make trending analysis results very timely and efficient.

Oil analysis interpreting is all about trend analysis and requires years of experience to get right. There is thus a natural synergy between AI and oil analysis.

AI will be able to spot and interpret trends.

My understanding is that AI is excellent at finding/comparing patterns to known (trained) data. This has an obvious use in trending oil analysis, which is essentially the science/art of identifying patterns in used oil data and interpreting them.

AI will improve analytics by being able to capture intuitive trends better than a human.
 
Idealistic approach would fetch new critical points and lines ensuring better results.
 
Consistent over trending.

AI will promote more extended service intervals by using tools like oil analysis.

AI can compare results and suggest possible causes of any abnormal variation.



Q.2. In what ways can AI be used as an ally in predicting degradation of lubricants in critical machines?

Use sensors-based monitoring in near real time and analyze the data to understand the trends that can impact oil life and/or asset integrity.

If it can indeed understand differing analytical responses for different base oil types, then it can potentially be predictive.
 
Oxidation.

AI can look at multiple data, simultaneously, to establish trends, and establish maintenance intervals and part replacement timelines.

If there was a publicly available database that AI platforms can access, with general data for specific machines, that would be ideal.

AI can reduce the amount of search time when researching cause and effect situations.

I don’t think AI is currently capable of interpreting results accurately enough to predict degradation and machine condition.

Potentially it can “see” trends and diagnoses more efficiently than human capability.

AI can be an ally because the responses can be quicker due to automation in testing and interpretation. It can also leverage historical data to make more accurate predictions about degradation.

AI should be able to review and trend the data faster than we can. It will be very critical that AI uses data specific to a type of lubricant, not general data about all lubricants however. If AI is taught details about a specific formula of lubricant it can make very prescriptive recommendations. If AI is using generalized data then it will be no better than today’s canned responses by testing facilities.

AI has one of its strengths in pattern recognition and as such can support staff in early predicting of (beginning) lubricant degradation.

AI can help increase the speed and accuracy of prediction.

AI could continuously watch trends and extrapolate to a predicted endpoint. Especially useful if the trend changes direction.

Extension of simple linear regression, build a more complex model to predict issues.

AI is a great tool in our toolbox of ways to sort through and correlate data faster and with greater accuracy. From that perspective, AI has the potential to save time and money for all oil analysis data users, especially the labs. The laboratories remain important in validating AI predictions. AI and sensors do not replace the lab; they redefine the role of the lab.

AI should be able to quickly adjust the results of a sample based on any minor changes in the equipment or input data. It may also allow for more data to be collected and sorted through such as environmental conditions at the time of sampling and during operation.

Detect trends faster, and see whether samples are “good” samples based upon any variances that don’t follow existing trends based upon all data available.

With the remote equipment installed, it can alert operations to issues well before the fluid becomes an issue.

I don’t think we have enough data to understand this. I mean certain models can be built. At the moment I do not have much faith in AI. Although there are some differences depending on which “engine” you use.

It may be able to identify some correlations not yet noticed by humans, but many of those may prove to be dead ends and cause time to be wasted.

Catch signals earlier, perhaps?

If it will be combined with data coming from different online sources, for example, machine temperature vibration and oil sensors/oil analysis.

Will AI (through sensors) report early lubricant degradation signs before labs?
Yes 85%
No 15%
Based on an informal poll sent to 15,000 TLT readers.
AI would record the minimal standard prescribed for a specific operation, their loads, temperature and duration of work to predict and signal possible failure or loss when the required attentions are not provided.

It can shift the highly subjective oil analysis to being objective!

AI could be used to analyze trends in oil chemistry change giving accurate predictions of useful oil life.

By giving more accurate analysis and indicating what may be causing the degradation (if abnormal). Through sensor use it could tell when oil needs to be changed rather than relying on interval changes.

It acts as a force multiplier for reliability teams. It handles the massive data crunching and pattern detection, allowing human experts to focus on strategic decision-making and addressing the complex problems the AI surfaces. It’s the difference between having a spreadsheet of past results and having a crystal ball for your machine’s health.

1.) For the analysis of historical data and finding solutions to overcome the issues. 2.) By checking the symptoms, causes and failures, it will provide a solution to avoid catastrophic failure, which eliminates the plant’s sudden shutdown. 3.) From the trending data, historical findings and elimination of defect prevention possible through an oil degradation identification mechanism through AI.

AI will eliminate the need for a lubrication expert. It will give information and make informed decisions for the company.

I think handling a lot of data is what AI should be used for.

Image analysis on the filter sometimes gives insight on the extent of degradation.

Oxidization analysis colorimetric measurements and debris particle classification have large potentials.

Machines can be equipped with sensors that reduce the need for taking samples.

Addressing historical oil analysis to identify repeating patterns in the data.

I do not know if AI will be able to predict lubrication degradation any better than the way that it is being done today.

Real time alerts through sensors (vibration/water ingress/temp) AI programs can combine the different results and report back in real time. This can be sent out through different platforms.

1.) Estimation through software. 2.) Evaluation after analysis approach. 3.) Autoprediction about result.

AI can be used in tracking the trend on oxidative thickening of the oil and on soot accumulation. This will help in oil drain extension programs.

If technology advancement happens ML sensors can detect oil degradation state and send information to the asset management via notification by automatic email,s which can help the maintenance team or technicians to plan their activity.

Recommends similar cases or solutions for recurring issues, and long-term evaluation of testing data to identify systematic deviations.

AI models can be built taking into account all operational parameters.

By using oil viscosity measuring sensors to monitor the contamination and degradation of the lubricant.

Especially in combination with new sensors, AI may be extremely useful to assess lubricant degradation inside of the application itself. Based on this information sampling may be reduced and planned in advance. Even after refilling the information collected during the previous fill will remain in the database and will be used to predict dates of sampling and refilling the current lubricant.

In combination with good sensors, AI will be a strong complement to actual measurements. Live analysis of large amounts of data from different types of sensor signals will make the quality of the predication and the possible response time better.

The AI software can be programmed to highlight projected results and the probability of lubricant failure continuously.

Advances in trending will bring better prediction models along.

Interrogating the life history of filed results and customer feedback against the interpreted results.

Earlier and more detailed results.

AI/ML can help to identify complicated interactions between monitored/sensor data that might be missed by a human analysis. The human should be in the loop because AI/ML can produce “artifacts” that may have nothing to do with what is being analyzed.

AI will have benchmark history data that can be analyzed by machine or component to improve predictive maintenance and reduce failures.

Digital twins could be created on existent data and be used to predict real time behavior.

Predictive analysis based on different parameters and the operating conditions analogy.
 
Daily interactions.

AI will use installed monitoring and predict an incident before it happens. Customers get to be warned sooner.

Editor’s Note: Sounding Board is based on an informal poll sent to 15,000 TLT readers. Views expressed are those of the respondents and do not reflect the opinions of the Society of Tribologists and Lubrication Engineers. STLE does not vouch for the technical accuracy of opinions expressed in Sounding Board, nor does inclusion of a comment represent an endorsement of the technology by STLE.