Skip to main content

30.06.2020 | Production + Production Technology | News | Onlineartikel

AI-Based Maintenance System Can Detect Even Unknown Damage

Thomas Siebel
1 Min. Lesedauer

An AI-based maintenance system for industrial plants can detect unknown errors in addition to known damage. According to the system inventors, novelty detection has not been possible with AI systems until now.

The system, developed at Saarland University, analyses sensor data from machines using artificial intelligence (AI). From the vast amounts of temperature and oscillation data, the AI system not only identifies known signal patterns indicating damage, wear and tear, or errors, but can also assign unknown errors to their causes, according to the scientists. This ability, known as novelty detection, is a first for AI systems, which until now have been unable to process unknown signal patterns.

When in use, this newly developed system constantly compares sensor data with normal values and patterns typical of emerging errors and damage. To achieve this, the working group selected a large variety of signal patterns that are associated with changes and damage to machines from a mass of measuring data. The scientists then trained their system using mathematical models for degrees of error. The program now continues to learn via automatic machine learning and can independently detect abnormalities. According to the scientists, the system's modular structure comprising hardware and software modules allows the system to be individually assembled to suit different industrial plants.

Weiterführende Themen

Die Hintergründe zu diesem Inhalt

2020 | Buch

Industrial AI

Applications with Sustainable Performance

Das könnte Sie auch interessieren

18.06.2019 | Manufacturing | News | Onlineartikel

Additive machines discover superalloys

01.10.2018 | Industry 4.0 | News | Onlineartikel

Sensor System for SMEs

Premium Partner


    Die im Laufe eines Jahres in der „adhäsion“ veröffentlichten Marktübersichten helfen Anwendern verschiedenster Branchen, sich einen gezielten Überblick über Lieferantenangebote zu verschaffen.