People see a limit and its a defined gospel; its religion for them, said John OMahony, oil laboratory supervisor for the lubricating oil department at Tullow, County Carlow, Ireland-based T.E. Laboratories, which offers oil monitoring services. Were trying to upset that and look at it in a different way.

These limits were developed in one region, with one climate, with one maintenance practice or procedure in place, OMahony stated. Its very hard to apply these limits to different environments and operating procedures.

OEM recommendations also dont consider varying levels of knowledge of component maintenance between companies and between countries, OMahony told attendees at the Society of Tribologists and Lubrication Engineers annual meeting in Atlanta.

You may also find this with customers using identical assets in one region. If they have different maintenance practices in place, youll get varied results in terms of your oil analysis data, he noted.

Green Means Go?

Oil monitoring can pinpoint particular issues with a lubricant, but OMahony observed that some of the laboratorys customers were reporting equipment failures despite following the traffic light recommendations-green, yellow and red to indicate the urgency of equipment maintenance-on their oil analysis reports.

We found some gaps where we werent highlighting potential issues that were happening, he noted. This led T.E. Laboratories to take a closer look at the data being collected from customers oil samples.

OEM recommendations, they found, should be viewed as guidelines rather than strict limits, because failures can occur even if an oil analysis report comes off as green, he explained. T.E. Labs instead uses big data to head off these problems.

T.E. Laboratories oil analysis lab evaluates samples of hydraulic fluids and engine, transmission and gear oils used in railway engines, stationary and mobile industrial units, generators, compressors and pumps. The 30 milliliter oil sample is collected onsite by the operator and then sent to the laboratory, along with forms detailing information about the oil and the equipment, to test for presence of wear metals, contaminants, additive levels, viscosity, fluid chemistry, fuel contamination and oxidative stability.

The resulting data allows the lab to anticipate which elements may cause problems and identify their root cause. The lab generates reports comparing the results of the analysis against defined limits, then sends them to the company to steer it toward preventive maintenance to avoid equipment breakdowns, which can result in costly repairs and unplanned downtime. The reports can be tracked by the customer in an online database.

The laboratory uses statistical model-based exploratory data analysis to determine chemical limits for the customer, which it says may differ from OEM limits. Currently, results are calculated using spreadsheets, but at the meeting in late May, OMahony said T.E. Labs aims to automate this process.

Three key steps for the statistical analysis are establishing valid data, defining subsets and identifying outliers. You have to be very careful in what [data] you use and how you use it, because that could have adverse effects when interpreting the results, he cautioned.

Subsets of data are established based on the customers equipment, such as a fleet of trains with a certain type of final drive. The final drive is the part of a vehicles transmission system that generates power and delivers torque to the drive wheels. That is one type of asset with very uniform operating conditions; it has the same oil, its the same asset, so that equipment is established as a subset of the customer itself, OMahony expanded.

Establishing outliers is important because some oil samples can vary in their formulation (for example, have an elevated presence of water) that can affect the oil analysis. We need to try and identify this and remove it from the data, because it can be very damaging to your overall result, he elaborated.

The Big Data that Could

T.E. Laboratories illustrated its recommendations through a case study of Northern Ireland Railways fleet of trains, which are manufactured by Beasain, Spain-headquartered CAF (Construcciones y Auxiliar de Ferrocarriles). The lab created four subsets for NI Railways fleet, gathering data for the diesel engines of the modern 4000 series trains, the aging 3000 series engines, as well as transmissions and final drives in both engine models, for a total of 2,236 units of equipment.

The laboratory analyzed the presence of wear metals, contaminants and additives in the oils. It also analyzed the sample oils viscosity, but OMahony warned that viscosity can be impacted by oil drain intervals and may not represent a definitive result.

For additives, OEMs usually do not have set limits because additive levels depend on the specific oil blend, OMahony noted. For a diagnostician, the important thing to identify is if its the right oil or the wrong oil within the asset.

CAFs recommended oil condemnation limits for wear-inducing contaminants on both types of trains were 60 parts per million of iron, 15 ppm of chromium, 10 ppm of aluminum and 20 ppm of lead, copper and tin. Despite heeding this, OMahony said, the rail operator was reporting failures from a brass journal bearing in the cam rod of the 4000 series engine, which the lab found was tied to copper accumulation.

Our reports were flagging green because the recommended limit was below 20 ppm, yet we still had failures, he reported. The laboratory subsequently brought down the copper limit to 6.5 ppm for the modern fleet after the initial analysis of the fleets data.

T.E. Laboratories set limits for the engine oils used in the 4000 series trains at 35.1 ppm of iron, 1.9 ppm of chromium, 4.1 ppm of aluminum, 3.8 ppm of lead and 1.1 ppm of tin, according to OMahonys presentation. For the aging 3000 series trains, the laboratory placed higher limits on the wear metals. Iron was set at 70.2 ppm, chromium at 3.4 ppm, aluminum and lead at 5.1 ppm, copper at 11.3 ppm and tin at 4.4 ppm.

The more modern fleet, as you can imagine, has a much shorter amount of hours on the equipment, said OMahony. With this in mind, the iron limit was set at a conservative 35.1 ppm to give the laboratory a good indication of the trains that are potentially going to relay issues in the future and the trains that wouldnt have any issues.

The laboratory also analyzed oils for contaminants such as barium, boron, silicone, sodium and water, among others. NI Railways was having problems with its seals, which T.E. Laboratories found was due to the amount of silicone in the engine oils. However, the information the operator had provided was too limited to pinpoint the source of the contaminant, and the lab solicited a clean oil sample from the rail companys supplier to compare against the aged oil.

Upon analyzing the virgin oil, we found out the root was actually silicone used as an additive within the virgin oil itself, said OMahony. After this discovery, the lab brought down the limit for silicone to 20 ppm for the older fleet and 18 ppm for the modern trains, versus the OEM recommended amount of 15 ppm, as a way to identify the silicone additive in the oil and set it apart from the silicone contaminant.

Since T. E. Labs began conducting data analysis on behalf of NI Railways last year, OMahony revealed that the rail company has managed to direct its resources more efficiently to address oil changes before failures occur. He added that the lab has been generating fewer potential failure reports than it did previously, when maintenance was done based on the OEM limits.

The key advantage with using this data analytics approach is the fact that you have a uniform system to diagnose any parameter, by using one standard deviation for warning levels and two standard deviations for failing, he said.

T.E. Laboratories data analysis has been applied to a number of its customers besides NI Railways, said OMahony. These reports and statistics are applied to their model, their procedures and their process, so its a very refined system for them. According to the labs website, customers include Irelands national bus and railway fleets, General Electric, alcoholic beverage producer Diageo and construction company Iron Planet, as well as shipping fleets, textile and injection mold industries, and mining companies.

With the continued influx of data, OMahony hopes that the system will self-regulate and recommendations can be made based on historical data. This system is dynamic, so it is constantly evolving. The advantage in this is that as a fleet ages over time, the limits may expand, but you are always highlighting the worst cases, he emphasized.