What is the risk of an unplanned downtime in your production line in the next couple of months?

What is the risk of delaying a preventive maintenance in order to meet a production commitment?

The Compound Anomaly Index, CAI, was created to help answer these questions. With predictive analytics  technology, it synthesizes how different your equipment measures are compared to when it was behaving normally, i.e., how anomalous is its behavior now.

In the following example, you can see CAI’s value for a cement mill over 18 months. It was calculated based on 300+ signals that came from 28 vibration sensors distributed over 8 distinct parts of the equipment. A machine learning algorithm used all theses signals to learn the normal behavior and predict the anomalies.


The CAI was zero for most of the time and started to change 130 days before the failure. A steep climb can be seen 100 days before the failure peaking at 71 days. It baked off after that but never return close to what it was when the operation was normal.

With that information in hand, two kind of actions could be taken to minimize downtime:

  1. Avoid preventive maintenance during the 214 days where CAI was at zero as the risk of a failure is very low.
  2. Intensify the number of preventive maintenance while the CAI is above 0 in order to avoid a bigger downtime to correct the failure and make the CAI return to zero.

CAI calculation can be done with several equipment with different types of sensors. It is neither specific to cement mills nor vibration sensors. CAI can be calculated for industrial metal press with metallic residues in lubrication sensors and power generator turbines with temperature, velocity, and current data.

We at Eleflow would love to hear what you think of this post so pleas leave a comment or contact us to learn more