Featured in COMPRESSORtech2, November 2017
A field demonstration of an online oil condition monitoring and auto-sampling IIoT device
BY CHAS OGDEN
It is common in the industry to use periodic analysis of lubricating oil with two fundamental objectives in mind. Firstly, this periodic analysis is used to gauge the suitability of the oil for continued use with the goal of performing condition-based oil changes. Secondly, best practices of oil analysis represent a powerful reliability tool and yields information about wear metals and contaminants, which may be used to identify mechanical issues and prevent catastrophic failures.
A key limitation of traditional oil analysis programs is the lag time from the collection of each sample (typically collected monthly), and the delivery of meaningful results by the lab (this lag time typically is seven to 45 days). Use of an automated oil sampling system for real-time oil condition monitoring offers advantages for operators. This paper explores how such a system was able to help understand and resolve oil fouling issues in a natural gas engine.
The real-time oil sensors measure metrics such as pressure, temperature, viscosity, dielectric constant, particle loading and flow rates for makeup oil (engine) and compressor lube oil (frame, cylinders and packing cases). Using data collection and communication systems, this data is transmitted to a proprietary “dashboard” where this real-time information and analytical results are available to secure web users.
The system features the ability to automatically fill oil sample bottles, which can be set for automatic sample collection either based on a routine operating time interval or based on a sample by exception. If the oil condition monitor’s edge processing yields a potentially dangerous trend in any measured parameter(s), such as viscosity, the system automatically fills a bottle and instantly notifies key personnel via email and SMS messaging. This sample is then processed by a laboratory with same-day turnaround so that results are available to the customer within days as opposed to one or more weeks.
Figure 1. Real-time oil condition monitoring system and and oil flowmeters mounted to test engine E1202 (left) and test compressor C1202 (right).
LogiLube LLC, the system manufacturer, and DCP undertook a field demonstration of the monitoring system on a key gas engine-driven propane refrigeration compressor in DCP’s East Texas Complex gas-processing facility outside of Carthage, Texas. Given that this machine plays a pivotal role in the production of liquid products in East Texas, it was expected that valuable information could be obtained that would help the reliability team at DCP ensure this asset’s availability by making real-time decisions about equipment condition and lube system performance.
Engine 1202 (E1202), a GE-Waukesha L-7044GSI engine, and Compressor 1202 (C1202), a two-stage Ariel JGK-4 compressor, were instrumented with the SmartOil systems for the duration of the 26-week test project (Figure 1).
Real-time oil quality measurements, oil consumption metrics, tangible machine events and laboratory analysis of physical oil samples are combined within a single database to provide data analysis and comparisons relating to an almost unlimited combination of parameters (i.e., particle count vs. number of shutdowns) and across different assets (i.e., oil viscosity of C1202 vs. E1202).
As a note, the two mechanical systems consisting of E1202 and C1202 share a common oil source. Given that the oil feeding both machines is guaranteed to be consistent, analytics can make comparisons relating to the effects of operating conditions (engine vs. compressor) on both the rate of oil degradation and the levels of measurement noise — interference — introduced into the real-time measurements.
Throughout the pilot, real-time oil quality and machine data was measured and recorded almost constantly throughout the 26-week project, interrupted by several small outages because of severe weather and/or unrelated power issues within the plant.
As part of the monitoring process, a real-time web dashboard is always available to the customer, allowing them to view data metrics and analytics results regarding their equipment instantly.
Data visualization — normalize certain measurements
The importance of communication and easy comparison of data highlights the need to provide data in clear and understandable ways. For the purposes of data visualization, metrics such as oil viscosity and oil dielectric constant are displayed in terms of percent change from a “nominal” value, which is the expected value of new oil at operating conditions. Empirical testing, or “characterization” of new oils allows the system to create mathematical models that remove the effects of temperature and other measurable properties, creating normalized trends displaying metrics such as viscosity and dielectric constant. Thus, the normalized percent departure from new oil can be illustrated.
Figure 2. Temperature normalized viscosity.
Figure 3. Temperature normalized dielectric constant.
Figures 2 and 3 detail the real-time normalized viscosity and dielectric constant data collected for both E1202 and C1202, where the two machines exhibit vastly different behaviors when compared. This observation is especially interesting considering both machines share an oil supply source and both experience the same environmental conditions. The steady-state nature of the C1202 frame oil properties indicate predictable and consistent operation. For the purposes of demonstrating the oil monitoring system, this observation is useful for two key reasons:
C1202 can be thought of as a comparative control subject with respect to E1202, which frequently exhibited nonsteady state and unexpected changes in oil properties. Since E1202 is an internal combustion engine and C1202 is a reciprocating compressor, it makes sense that the compressor frame oil circulating will not be exposed to the chemically active combustion gasses, fuel or temperatures within E1202.
Using C1202 as a control subject, it was possible to separate the effects of environment and variations in the supply oil from operational considerations.
Abnormalities — dielectric constant
Upon commissioning, the E1202 oil monitoring unit immediately alerted that the dielectric constant exceeded the maximum or allowable limit previously identified for that oil type.
It was proposed that the monitoring unit had either been incorrectly calibrated or programmed with an incorrect oil baseline. Given the unexpected initial measurements, these theories seemed plausible. There were, however, factors indicating otherwise. For one, viscosity had also started in the “warning” zone established per SAE grades, between the upper edge of “nominal” and the “danger” limit.
To determine the cause of the dielectric constant “danger” alarm and potentially identify corrective action, oil samples were immediately sent to both the oil analysis lab — where the oil was analyzed for clues to explain the high dielectric constant — and LogiLube’s R&D facility, where a second SmartOil sensor array measured the oil to corroborate the measurements.
Figure 4. Water content comparisons for E1202 (left) and C1202 (right).
Results from the lab indicated that the oil had a notable concentration of water (Figure 4) (206 ppm, common for a rich-burn engine) and elevated oxidation and nitration. Meanwhile, lab measurements confirmed the E1202 unit was reporting data correctly, where the average difference between lab- and field-measured dielectric constant was <0.6%. Despite the notable concentration of water in the E1202 oil sample, this level of water contamination is not sufficient to cause high dielectric constant “danger” reading on its own. Upon validation of the dielectric constant measurements, the problem was studied in-depth to understand its possible cause.
In a broad sense, the initial dielectric constant “danger” alarm on E1202 and subsequent upward trend begged the question, “Why is the oil in E1202 becoming more polar over time?”
One month into the project, the basic differences between oil measured from E1202 and C1202 became clear: consistently higher iron, water, total acid number (TAN), total base number (TBN), oxidation and nitration levels were found in E1202. Copper, however, was higher in C1202, and the C1202 metrics all appeared to be relatively constant, whereas metrics related to E1202 were increasing over time.
To better understand the underlying effects of an internal combustion engine on lubricating oil, it is important to discuss how oils break down over time within an engine. Oil degradation is caused by heat, mechanical wear and contamination by combustion byproducts. Heat and contamination by combustion byproducts typically contribute to chemical degradation of the oil while mechanical wear typically contributes to physical compromise. Regardless of the type of breakdown, the result is the same: The oil loses its ability to perform its intended functions, and the engine is more likely to suffer from decreased reliability and longevity.
Oil breakdown manifests itself in the following ways:
Depletion of oil additives (i.e., antioxidants, antifoam, detergents).
Reaction of atmospheric nitrogen with oil base stock (nitration).
Reaction of oxygen with oil base stock (oxidation).
Precipitation of inorganic solids (ash formation).
Sludge/varnish formation (accompanied by viscosity increase).
The following hypothesis was developed with the goal of relating oil degradation back to dielectric constant:
Since the byproducts of chemical breakdown in the oil are polar, continued oil degradation will result in an increased polar species, meaning that over time the oil would become more polar.
Once a clear hypothesis was developed linking the E1202 dielectric constant measurements to aspects of oil degradation, evidence would have to be gathered to support (or disprove) the hypothesis. Ideally, datasets supporting this theory should be able to indicate a marked change in oil parameters driven by a controllable event to remove a maximum number of confounding elements and assist in identifying solid cause-and-effect relationships between oil degradation and dielectric constant measurements.
The solution identified to provide the necessary datasets for the evaluation of this theory was to observe the oil properties both before and after oil changes that were conducted on the compressor package. At the request of the vendor, an oil change was performed at the end of January 2017, four weeks into the pilot project. This first oil change was intended to bring the oil properties of E1202 back to a known starting point so that the cause of the dielectric constant “danger” alarm could be pinpointed.
It was assumed that an oil change could be approximated as a total oil changeout, effectively replacing all used oil with new oil, which would be expected to correspond very closely to the values reported by the compressor, C1202. As a result, the data analysis team could evaluate the effects of replacing the oil within E1202 with “new” oil, allowing for a before-and-after comparison.
Figure 5. Oil change performance.
The first oil change produced results that initially con- founded the data analysis team. Despite the oil change, the oil monitoring unit installed on E1202 provided oil parameters that did not reflect the measurements reported by C1202. The real-time dielectric constant remained elevated above the “warning” limit and real-time viscosity measurements also reflected higher viscosities than the expected C1202 values (Figure 5).
Initially, these results appeared to disprove the theory that oil degradation is linked to the dielectric constant measurements. However, an intriguing trend began to emerge. None of the oil properties decreased to expected levels after the first oil change. The changes in oil properties were calculated from averaging the analysis results from the three samples before and after the first and second oil changes. The change in properties in Figure 5 is displayed as both a numerical difference and a percentage difference, so that a 100% drop would mean that the oil properties had changed to represent those of “clean/new” oil. The small difference in oil properties after the first oil change indicated that the oil monitor measurements agreed with the lab, providing justification for the sensor data and indicating the oil change was somehow unable to refresh the oil within E1202.
The simplest explanation for the poor performance of the first oil change was that the oil was not completely changed, in fact a substantial amount of residual oil must have been left in the engine, given the small improvement in wear metal concentration and chemical properties. The customer was approached for a review of their oil change practices to identify the cause of the residual oil, but the customer was adamant that it had completely and thoroughly drained the engine oil.
While convenient, an incomplete oil changeout was not a satisfactory explanation for the chemical properties of the oil, most notably oxidation, nitration, TAN and TBN. These values failed to decrease as much compared to physical contaminants while oxidation even increased. It became clear that further investigation was required into the effects of chemical degradation on the oil.
Nitration occurs when engine oil becomes contaminated with nitrogen oxide compounds, usually introduced into the engine via the combustion process, exhaust gasses, etc. Nitration represents a form of oil breakdown and can be exacerbated by conditions such as poorly functioning piston rings, fuel-air mixture problems and improper crankcase ventilation. Nitration is typically most significant in cases of low-speed, four-stroke engines, whereas two-stroke and high-speed, four-stroke engines are less sensitive. According to ALS Tribology, “an excessive level of nitration in a natural gas engine or compressor application is typically accompanied by an increase in acid number and viscosity.” Such trends in TAN and viscosity were readily visible in the datasets provided by the analysis lab and the oil monitoring system, further supporting the theory that dielectric constant was indicating oil degradation.
Figure 6. Dielectric constant versus wear metals.
Oxidation occurs when reactive oxygen species are created in the high-temperature and high-pressure environment of the cylinder of an internal combustion engine. These reactive oxygen atoms (free radicals) interact with the lubricating oil lining the sides of the cylinder wall and cause it to break down. Sacrificial antioxidant additives (such as hindered phenols) are added to the oil, neutralizing the free radicals that damage the properties of the oil.
Oil degradation byproducts include alcohols, ketones, aldehydes, water and peroxides, all of which are highly polar compounds that increase in concentration over time. It is important to note that deposits such as varnish and sludge that accumulate in the inner surfaces of machinery are known to be chemically active and shorten the life of new oil added to the system during a top-up or change. This could help explain why the team saw results similar in appearance to an incomplete changeout of E1202’s oil.
After the second oil change, a staggering improvement was visible across all oil metrics, physical contaminants, chemical properties and real-time measurements. While this likely indicated a process improvement by the customer, ensuring that the old oil was completely drained, a process improvement was unlikely to produce such a profound difference, especially given the customer was certain about draining all the oil. The second oil change provided strong support for the theory that chemically active deposits were left in the engine after the first oil change, creating an environment where sacrificial additives in the new oil were immediately depleted while attempting to neutralize the varnishes and sludge left behind. In essence, it cleaned the engine.
Figure 7. Slope change performance.
While it is tempting to assume that wear-metal content and dielectric constant must be linked (because metals are excellent conductors), Figure 6 indicates that this is not the case. Wear metals are not dissolved in the oil; rather, they are solid particles suspended in the lube oil stream. They do not impact the chemical properties of the oil and thereby do not affect its dielectric constant.
This does not detract from the relationship that exists between dielectric constant and wear metals demonstrated on E1202. Dielectric constant, iron and copper are highly correlated, confirming that the wear metals are somehow linked. The largest common factor between wear metal content and chemical degradation of the oil (indicated by dielectric constant) is oil operating age where, assuming normal operation, the oil is expected to accumulate wear metals and suffer chemical degradation as it circulates throughout the engine.
While this appears to be the case for E1202, it is also possible that the chemical degradation of the oil contributed to presence of wear metals through reduced lubricant performance and engine protection.
Degradation rates and contamination rates are detailed in Figure 7.
After the first oil change, wear rates of iron and copper decreased substantially along with oxidation, TAN and TBN. The dielectric constant rate of increase also slowed, but not as much as the other parameters. Nitration and viscosity rates increased, which may support the assertion that the new, fresh oil was subjected to immediate chemical degradation via residual compounds within the engine (Figures 7, 8).
Figure 8. E1202 dielectric versus lab nitration, oxidation (left) and E1202 viscosity versus lab nitration, oxidation (right).
After the second oil change, most parameters illustrated a sharp decrease in rate of change, indicating that the oil properties appear to have become more stable. Notable exceptions are iron, TAN (the cause of which is under investigation) and TBN (a positive indication since a higher TBN value is indicative of less oil degradation).
While the measurements after the second oil change indicated that the oil became more chemically stable and that there was a general reduction of wear rates, the behavior of oil properties was seen to drastically change in the middle of June. Denoted in the graphs as event five, this behavior change marks the point at which chemical degradation and contamination effectively flatline, or in some cases, begin to decrease. Note the alignment of event five with respect to oxidation, nitration and daily oil consumption.
Did oil consumption cause the leveling of oil properties or did the oil properties increase oil consumption?
The competing theories developed to explain the behavior change lead to a chicken-and-egg type of problem. In the first theory, the fresh oil introduced after the second oil change was able to clean the internals of the engine, sufficiently reducing sludge and varnish deposits, allowing oil to more easily flow into valve guides and cylinder walls — contributing to the higher oil consumption. In the second theory, a leak or other unnamed source of oil consumption lead to an increased dilution rate of crankcase oil within E1202. This would help explain the decelerated rates of change exhibited by the oil properties in question.
Figure 9. Engine daily oil consumption versus oxidation and nitration.
While the causal relationships between oil consumption and event five are still being identified, the increased oil consumption rate from 0.423 gpd (1.6 L/d) to 1.753 gpd (6.6 L/d) in E1202 constitutes roughly a 1.4x increase in the dilution rate of the oil sump. That means that over a long period of time, the in-service oil is being diluted with new oil 1.4x as fast, which also means that the new oil can counteract contamination by the same multiplier. The marked decrease in all indicators of chemical degradation and mechanical contamination suggests that dilution due to oil consumption may be a contributor to the behavior change; however, it is not the sole cause (Figure 9).
While further study is required to address this behavior change with any degree of certainty, the ability of the SmartOil to detect and characterize such phenomena highlights its application as a powerful tool for prognostics and diagnostics of high value machinery.
Today, the oil monitoring system and real-time dashboard allow managers at the site to properly time oil changes and schedule maintenance based intelligently on awareness of current machine conditions. This reduces the risk of maintenance-related failures, provides peace of mind and ensures that high-value assets are serviced timely and appropriately without incurring unnecessary downtime.
The monitoring system and data analytics services operate on the idea that oil analysis is improved — not replaced — by the ability to measure oil properties and consumption in real time. The real-time sensor measurements, automated sampling and analysis results produce a picture of overall lubricant and machine health consisting of over 56 million data points per year.
The rate of change in dielectric constant before and after an oil change has a strong correlation with oxidation and nitration levels observed in lab oil analysis. At this stage, it can be preliminarily concluded that, elevated levels of nitration and oxidation in a rich-burn engine can cause additive depletion which, in turn, can increase the dielectric constant of in-use oil, allowing for the real-time calculation of the remaining useful life of the lubricant and the early detection of harmful machine operating conditions. Further assimilation of real-time and lab analysis data is required to corroborate and establish a strong analytical model between these variables, building on the findings of this 26-week demonstration project.