top of page
Home: Welcome
turbit-04-trans.png

Arbeiten mit
Turbit KI

Minderleistungen erkennen, untersuchen und frühzeitig beheben. Abnormale Verhaltensmuster von Komponenten werden sofort entdeckt. Durch das Feedback unserer KundInnen wird die KI von Turbit stetig besser.

Home: Headliner
  • How much data does Turbit need in order to work?
    For the training of a module of turbit we ideally need 24 months of data. If less data is present we can use transfer learning and start with one months of data. You can read here what kind of data you need for each module to work.
  • How does the Turbit monitoring work?
    Turbit is based on artificial neural networks, machine learning, and modern data engineering. You can read more about how Turbit works here.
  • How many alarms do I get from Turbit?
    Turbit has a very low false positive rate (below 10%) and you can manage about 300 turbines with 30 minutes per week. Depending on the status and quality of your fleet you can get 5 alarms per 100 turbines per week.
  • Do I need a software training to be able to work with Turbit?
    No, Turbit is very easy to use and you should be able to understand everything at the first use. On the EventCard every analysis is described and easy to understand. However, we give you an in depth workshop at the beginning and continous support in weekly or bi-weekly customer success workshops with our experts.
  • What does Turbit monitor?
    Turbit monitors the power and the main components of the turbine. You can find more details here.
  • How is Turbit's technology different from other vendors?
    Turbit monitoring works in real-time as well as fully automatic. Turbit has a failure mode prediction powered by AI.
  • How long does the training of a neural network take?
    The training of the neural network takes less than a day. Once we have all data present and all checks are green we are online for your whole fleet in less than a week.
  • What type of failures does Turbit detect?
    The answer highly depends on the data quantity and quality of your turbines data. Before we start a collaboration we do a preliminary data check to ensure maximum coverage. Turbit detects more than 35 different types of failures from the unknown power curtailments or power curve problems to a clogged oil filter or a main bearing problem that needs a complete exchange.
  • Do you need to install hardware at the turbine?
    No, Turbit is only based on software and works very well with SCADA data, there is no need for additional hardware. We can, however, use additional CMS or other sensor data like blade vibrations to get even better results and higher failure coverage.
  • How does Turbit ensure that you have the right signal mapping?
    Turbit uses Data Analytics and Language Models as well as manual checks to calculate a similarity score of signals or in other words a mapping probability. If these probabilities are off we perform manual checks and correct the signals. Often, signal mappings are changed over time, Turbit then detects an anomaly and a wrong mapping or frozen sensor signal can automatically be detected.
bottom of page