Real-Time Oil Analysis Has Arrived
EP Editorial Staff | February 12, 2020
Sensor technology is changing the way we approach oil analysis and will alter the role of labs in the near future.
By Mark Barnes, PhD, CMRP, Des-Case Corp.
For any critical oil-lubricated asset, oil analysis is a vital predictive-maintenance and proactive tool. Unlike other condition-monitoring tools, such as vibration analysis and thermography, that predominantly focus on detecting the onset of an incipient failure, oil analysis is capable of determining the causative factors that can lead to a failure such as the wrong lubricant, degraded lubricant, or contamination from particles or moisture. It is this proactive aspect of oil analysis that makes it such an indispensable tool in the condition-monitoring toolbox.
Traditionally, oil analysis has been route based, just like vibration analysis. Unlike vibration analysis, where a trained analyst can immediately identify a problem by looking at characteristic defect frequencies or even the time waveform, oil analysis requires the additional step of submitting a sample to an oil-analysis lab—either onsite or offsite—so that detailed physical or chemical tests can be conducted to try to identify issues. This additional step can often add days or, in extreme cases, weeks to determining whether there is a pending problem. While this is not always an issue in slower moving equipment, in higher speed machinery or mobile-fleet applications, any delay can be catastrophic.
One thing that vibration and oil analysis share is the nature of the result obtained. Whether it is a vibration analyst making a reading or a lubrication technician acquiring a sample, both technologies are taking an instantaneous snapshot of what is happening with the machine at the moment of the reading or sampling. In situations where machine operation is constant, with the same load, duty cycle, and operating parameters, this may seem like a subtle point. However, for assets where operational conditions are constantly changing, the results can render either technologies less effective.
By way of illustration, consider a vibration analyst taking a reading from an electric motor driving a centrifugal pump. While the load on the pump and motor might be constant, if the pump is operating close to a vibratory screen that runs intermittently, the reading from the motor may differ depending on whether the screen is running or idle. In vibration analysis, a good analyst is trained to account for operational variants. In oil analysis, since the analysis is one step removed from sampling, operational changes may go unnoticed.
In oil analysis, the data is also an instantaneous snapshot of how a machine is operating at the exact moment the sample is acquired. This can have a profound effect on the results. Consider the example shown in Fig. 1. The data shows the results of flushing a hydraulic system using a bypass filtration unit during normal operation. To insure that filtration is having the desired effect, the overall cleanliness of the oil is being monitored using real-time particle counting in the 4-, 6-, and 14-μm particle size range per ISO 4406:99. While the overall trend in particle contamination is decreasing, as measured by the linear 4-μm trend, the shot-to-shot variation in all three particle size ranges is significant, corresponding to when pumps, valve, and actuators are starting and stopping.
Now, imagine taking a bottle sample sometime during the trend shown in Figure 1. Depending on the machine-operation phase, it is highly possible that the sample will by taken at either a peak or trough in the particle-count trends. As such, the reported particle count may appear erroneously high or low since the bottle snap is simply an instantaneous snapshot of what is happening in the system. Only by monitoring particle count in real time can the true effectiveness of the bypass system be determined.
Figure 2 shows a similar example. In this case, a “smart”-desiccant breather is monitoring the relative humidity in the headspace of the reservoir of a hydraulic system. Based on a bottle oil sample taken during normal operation, the water content in the oil was measured as 30 ppm (0.003% v/v), well below the limit for this application. However, two days after the sample was taken, the operations crew elected to clean the system using pressurized water. Based on the real-time data from the sensor, the headspace, and hence the breather, saturated within 24 hr. This alerted the maintenance team, in real time, about a potential catastrophic situation that could have resulted in hydraulic-pump cavitation or other issues. The next routinely scheduled oil sample was 25 days out, meaning that this problem could have continued for another three or four weeks without being identified.
In addition to particle-count sensors and smart-desiccant breathers that measure the degree and direction of breather saturation in real time, the number and type of sensors available for monitoring lubricants has grown dramatically in the past five years. Today almost every physical or chemical property of oil that has traditionally been monitored using lab-based oil analysis can be monitored in real time using a sensor. The table above compares the most common oil-analysis lab tests with readily available sensors that can emulate the changes in various oil physical and chemical properties. As the table indicates, the only oil-analysis parameter that cannot be easily replicated in real time is elemental analysis that relies on atomic emission or absorption to report the concentration of wear metals, certain contaminants, and inorganic additives.
With many sensors to choose from, the most obvious question is: Which are most effective? The key to answering that question is to thoroughly understand failure modes and how quickly certain oil properties change with time. While some homogeneous properties of oil, such as viscosity, change much slower than others. Information that can change daily, hourly, every minute, or even every second, is far more valuable to monitor in real time. These include particle and water contamination, wear debris content, and, to some extent, overall oil quality. Comparing the constantly changing parameters with real-time operational parameters such as load, speed, temperature, operation cycle, and execution of certain maintenance procedures can provide unprecedented insight into how the lubricant is responding to normal and abnormal operating conditions.
Like all sensor-based condition monitoring technologies, real-time oil analysis is not intended to replace bottle samples and lab analysis. For the short to medium term, there will always be a need to trust the knowledge, skills, accuracy, and precision offered by a commercial oil-analysis lab. In the future, oil-analysis labs will become more like “forensic labs,” using their expertise to determine the underlying root causes and reasons for problem conditions first identified using sensors such as lubricant degradation or increased wear trends. Sensors will become the de-facto warning to check oil or machine condition. Combined with operational data, we are entering a new era in which integrated condition monitoring and operational context will give us unprecedented insight into how our equipment functions. Combined with “big data” machine learning capabilities, these are exciting times for oil analysis. EP
Mark Barnes, CMRP, is Senior Vice President at Des-Case Corp., Goodlettsville, TN (descase.com). He has 21 years of experience in lubrication management, oil analysis, and contamination control and has published more than 150 technical articles and white papers.