Oct 15, 2018 in Uncategorized

Our Predictive Asset Management (PAM) project at one of the largest breweries in the Netherlands has once again demonstrated the power of condition monitoring and data-driven modelling. Recently it has been determined that the end of a clear trend and peak in one of the servo drive data points corresponded to the replacement of a bad knife of the foil system in the machine.

When the condition of a knife in the foil station deteriorates, it has major consequences for the entire packaging line and usually leads to unplanned downtime with production loss as a result. Not only the repair itself, but also the many short stops in the run-up to that, have left their mark on the production speed, as well as the rejected products. In this case there was a total production loss of approximately 100,000 articles.

The upward trend that we saw after our signal processing was abruptly interrupted when the knife was replaced. This is again a clear fingerprint in the data, this time from a knife that has become blunt. We can predict a deterioration of the knife with some confidence, so that it can be replaced in time and on schedule.

This is yet another good example of what Machine Learning methods can achieve within an (industrial) maintenance organization, where countless data sources are already available. With the help of our data infrastructure, algorithms, data visualisations and the knowledge of the maintenance engineers, this result was finally achieved that underlines the enormous potential of Predictive Maintenance.

Also read our article on bearing failures .