Asset Failure is Unique

Asset Health Needs More Than Reactive Maintenance

Are all machine breakdowns alike? The Anna Karenina Principle says no, and sheds light on the shortcomings of a strictly reactive approach to asset health. Reinforce existing maintenance solutions with PdM technologies that use data anomaly detection to deliver actionable asset health insights.

Leo Tolstoy begins his epic 1878 novel, Anna Karenina, with the following statement: “Happy families are all alike; every unhappy family is unhappy in its own way.” Though it may have surprised Tolstoy, this so-called Anna Karenina principle has since been applied to business strategy, personal relationships and even statistical modeling. The underlying argument is that systems that work perfectly resemble each other, but each instance of failure is unique.  

The Anna Karenina principle can similarly be applied to Big Data and is of particular relevance for Industrial IoT (IIoT) Predictive Maintenance. Today, the frequency of machine breakdown on the floor (i.e., working machines becoming malfunctioning machines) is likely determined by the type of monitoring being applied to the Big Data they amass.

The Industrial Machine Family Archetypes

Industrial machinery is comprised of multiple interconnecting components with embedded sensors. Data scientists use the data that is generated from these sensors to monitor systems and predict when failure is likely to occur.  

 

Machines that work without fault are the “happy families” – they seem identical and show no impending sign of breakdown. However, when the family is “unhappy” or the machine breaks down, each failure is unique and difficult to predict. 

SCADA Data Monitoring is Too Little, Too late

A traditional assumption within machinery asset maintenance is that past patterns of failure can be used to predict future incidents. The famed philosopher George Santayana stated that “those who cannot remember the past are condemned to repeat it, and many manufacturers seem to have had this saying in mind when crafting their maintenance strategies.  

For instance, plant technicians monitor SCADA (supervisory control and data acquisition) data and are alerted as to whether sensors have breached manually-set control thresholds. Based on past experience, if these thresholds are breached, it is indicative of a fault in a machine or system. The problem with monitoring SCADA data is that by the time a threshold has been breached it is already too late. It is easy to identify the “unhappy family” in front of a family court, but at this point, remediation is an unlikely occurrence. 

Supervised Machine Learning Doesn’t Accept Anna Karenina’s Premise 

Within data science, there are a number of approaches to Machine Learning. In Supervised Machine Learning, the data is labeled with examples of failure and is trained to identify new incidents of failure based on these labels. In our analogy, the model is trained to recognize a new unhappy family because it resembles other unhappy families. 

To apply this to Predictive Asset Maintenance, if the Machine Learning algorithm can recognize common failure patterns, then industrial plants can be warned about a potential failure before it occurs.  

Herein lies the problem. Because each “unhappy family” is different from each other, there are not enough examples to train the algorithm for each type of failure. As a result, the Supervised Machine Learning algorithm lacks sufficient data to be trained and cannot detect an evolving failure. 

Anomaly Detection Bypasses the Similar/Unique Binary 

An alternative that we use at SKF Enlight AI is based on Anomaly Detection. We do not assume that each machine failure will be similar. Instead, we are looking for anomalous patterns. To use the Anna Karenina principle, we are looking for behavior patterns that are unusual as a way to predict evolving asset failure.  

In a sense, anomaly detection provides a deconstructed version of the dichotomy Tolstoy presents the readers of Anna Karenina with: “Yes, all happy families may seem alike; however, this is a misconception. if we closely observe the way these families act on a daily basis we discover that they may already be somewhat unhappy.”  

To abandon the metaphor, just because a machine is producing at the same rate it was producing yesterday does not mean that it is in pristine condition. By evaluating the ongoing data streamed from the machine sensors and using healthy machine behavior as a model we can identify erratic machine behavior long before it reaches breaking point. 

Adopt Predictive Maintenance Strategies with Anna Karenina in Mind 

The Anna Karenina Principle provides a working, if not perfect, analogy for the use of historical data for failure prediction. On a macro level, it is a useful way of framing the problematic use of Supervised Machine Learning methodologies for downtime prediction. If we only look for what we can recognize – the unhappy family based on simplistic criteria – then we miss the opportunity to resolve what we cannot yet see: early signs of evolving degradation and breakdown. 

 

Leave a Reply

Your email address will not be published. Required fields are marked *

*