Researchers from Germany’s Saarland University have developed a maintenance system that combines artificial intelligence (AI) and sensors to monitor the status of industrial equipment and machinery.

According to its developers, the system can reportedly detect sings of damage, wear and error states based on data culled from real-time sensors and regular industrial machine sensors. Under normal operating conditions, industrial machines emit their own unique signals — including vibrations, hums and shakes — and changes in those signals serve as early warning signs of possible wear or impending failure.

Constantly comparing these signals against data culled from real-time sensors and the normal operation of the machine enables the system to detect any differences in the signal patterns that could be associated with wear, error or failure. Once detected, the system alerts machine operators and instructs the operator how to respond to the issue.

Additionally, the system can also recognize signals not previously emitted by the machines. This is accomplished via AI, which functions via pattern recognition. In the event that a new, unrecognized behavior or signal occurs, the system notifies its operators. This is a process dubbed “novelty detection” wherein the team detects a new signal, learns from it and then assigns the behavior’s cause to an associated source.