Predictive maintenance is one of the pillars of Industry 4.0 and a key driver in its digital transformation. Adopting it means choosing a proactive strategy instead of a reactive one, i.e. to transform or die in the digital age. But, what exactly is predictive maintenance and what does it entail?

Predictive maintenance is a technique employing machine learning technology, and it is used for assessing variations in an asset or a machine’s normal behavior, or even the behavior of its environment, in order to detect possible anomalies and predict future equipment failures before they arise.

Its monitoring makes it possible to collect data on changes that may take place in their behavior and analyze them. It’s primarily based on the following methods:

  • An analysis of the causes of an incident in order to understand why it occurs, what brought it about, and how to prevent it.
  • The detection of anomalies through an analysis of the symptoms and the probability of them reoccurring.
  • Maintenance system planning through the use of data logging to make it possible to predict incidents over the long run.

The goal is both to cut down on the amount of failures as well as the impact they could have, which usually entails service interruptions and, as a result, financial losses; in addition to repercussions on the offered level of service, allowing the necessary corrective adjustments to be performed without bringing the productive cycle to a halt. That is how the performance of unneeded maintenance tasks can be avoided. For example, this is the main difference compared to more traditional preventative maintenance solutions, which program regular maintenance tasks, regardless of whether they’re needed or not, in addition to collecting and analyzing data manually, giving rise to wasted time and resources.

What technologies are employed in the practice of predictive maintenance?

Predictive maintenance is a solution based primarily on IoT, advanced analytics and artificial intelligence.

IoT: connected devices outputting data captured on how they are functioning.

Advanced analytics: both current and past functioning, which lets you find out how the variables normally function and calculate the viability risk for the behavior established as “normal”.

Artificial intelligence: automatically learns the cause and effect relationship of the variables that characterize the asset’s behavior. This allows predictions to be made on functioning according to the changes that occur in the analyzed variables, thus permitting automated decision-making based on the log data. Data models continuously optimize the predictive algorithm in order to best respond to performance and maintenance problems.

How is predictive maintenance used in different sectors?

One example is the Spanish energy company Cepsa, which has deployed a predictive system in all of its Andalusian refineries. Via data ingestion on its equipment sensors that measure variables such as pressure, temperature, and power, the automated learning models are trained to accurately predict and anticipate possible anomalies. This way future faults in the machinery can be avoided, performance can be measured in real time, false alarms can be detected and downtime can be avoided.

The German automotive company Audi has also implemented a 4.0 maintenance plan in one of its body shops located in Neckarsulm (Germany). Its assembly machines have a punch-riveting system which uses compressed air to drive the rivet punch through a tube, causing wear on it. Through analyzing millions of pieces of data from sensors installed on these machines, it’s possible to minimize unexpected equipment failures and perform maintenance tasks without interrupting the production cycle.

Implementing this kind of system means increasing the reliability and availability of industrial assets. Avoiding unplanned stoppages, outages and downtime during repairs extends component service life, reducing the stock of spare parts and the expenditure on insurance. These are just some of the advantages of deploying such a system. As a result, the right predictive maintenance strategy involves increasing productivity, and saving time and money.