Predictive Maintenance & Anomaly Detection

Increase availability through the Sentian AD-PdM

In few areas is it more important to be leading edge. Our customers have important processes that need to be available 24/7 e.g. powering a city with heating and electricity.

Sentian’s solution is a deep learning and multi AI advanced solution for predictions and anomaly detection.

Using the latest in AI we can find more anomalies and make better predictions. The solution is also fast to implement, learns quickly and will offer fast ROI.

The Sentian Predictive Maintenance and Anomaly Detection solution covers both unsupervised and supervised conditions. It combines proven AI approaches with the latest in deep learning. Sentian employs Automatic ML to speed up development and changes as well as to efficiently adapt to varying local conditions, both in the field and on the factory floor.

• Increase availability of machines

• Discover problems early

• Reduce cost of maintenance

• Optimize maintenance

• Plant wide scaling

• Root cause analysis

• Cascading problems detection

• Learning anomaly detection

Predictive Maintenance for complex use cases

Predictive maintenance usually cannot detect cascading problem patterns.

We have created a solution that understands the machine/plant structure and combines it with advanced AI to find complex patterns over time.

Anomaly Detection for robust machines

Modern anomaly detection in large scale industrial implementations can generate lots of false positives.

We solved this by creating an anomaly detection system that works with local domain experts and integrates with control systems. It also learns and improves over time, ensuring long-lasting results.


• Advanced pattern recognition

• Learning anomaly detection

• Cloud based and on-prem

• Supervised and unsupervised learning

• Topology optimization

Anomaly Detection, Predictive or Prescriptive Maintenance

There has been a move from talking about Predictive or even Prescriptive Maintenance to Anomaly Detection as industrial companies realize that their data may not be good enough for predictions. This may be from a lack of labeled data1 or that the machines are simply too robust for predictions, i.e. do not break often enough for patterns to emerge.

We see improved maintenance as a continuum and have built solutions that span from unsupervised anomaly detection to predictive maintenance all the way to optimized maintenance, also called prescriptive maintenance. This makes it possible to evolve the solution from anomaly detection to a predictive solution over time as the data matures.

1Labeled data is data that can serve as an example for the AI to learn from, e.g. data that is linked to maintenance events.

Want to learn more?