{"id":1920,"date":"2018-11-22T06:36:05","date_gmt":"2018-11-22T11:36:05","guid":{"rendered":"https:\/\/blogs.plm.automation.siemens.com\/t5\/Simcenter-Blog\/Neural-networks-amp-digital-twins-change-the-O-amp-M-in-the-wind\/ba-p\/544642"},"modified":"2026-03-26T05:53:38","modified_gmt":"2026-03-26T09:53:38","slug":"neural-networks-digital-twins-change-the-om-in-the-wind-industry","status":"publish","type":"post","link":"https:\/\/blogs.sw.siemens.com\/simcenter\/neural-networks-digital-twins-change-the-om-in-the-wind-industry\/","title":{"rendered":"Neural networks & digital twins change the O&M in the wind industry"},"content":{"rendered":"

Today wind power represents 4.4% of the total generated power. By 2030, this is to increase up to 20%. The challenges for wind turbine manufacturers are wide-ranging: the aerodynamic performance of the blades, reduce weight, keep noise and vibration levels under control, ensure a durable design and improve its overall system performance.<\/p>\n

The gearbox is the most critical part of the wind turbine. Either you send a technician up the turbine and do a manual check, or you attach sensors to the gearbox and monitor the results remotely on a computer. Both approaches work to anticipate failures and allow turbine owners to schedule for repairs. Obviously, this comes at a price. A high price. Can\u2019t this be done more cost-effective?<\/p>\n

Predicting the remaining useful lifetime of each wind turbine gearbox<\/strong><\/p>\n

Winergy<\/a>, a global key provider for wind energy in Germany, teamed up with the Simcenter Engineering\u00a0experts of Siemens PLM Software to estimate the remaining useful lifetime (RUL) of a complete wind park. Let\u2019s be a bit more specific: 78 wind turbines \u2013 35 SCADA channels \u2013 historical data stored over 4 years.<\/p>\n

The Simcenter Engineering<\/a> specialists tackled this issue by combining 2 approaches:<\/p>\n

    \n
  1. \n
      \n
    1. Neural Networks<\/strong>
      \nThe neural network was fed with information from different SCADA channels on the gearbox in combination with service data. Gearbox temperatures were defined as the most representative signals for a possible failure. Next, the neural network was trained on how a turbine reacts in healthy and faulty conditions. Winergy and Simcenter experts used the technique to accurately predict and detect failures early on.<\/li>\n<\/ol>\n<\/li>\n<\/ol>\n
        \n
      1. \n
          \n
        1. Digital Twin<\/strong>
          \nA digital twin makes the bridge between a virtual representation and the physical product. It helps to understand and predict product performance characteristics. Wind turbine modeling was combined with physical validation measurements in 1 turbine to validate the digital twin model. The digital twin model is fed with historic loads extracted from the SCADA in order to predict the remaining useful lifetime of the bearings and gear teeth in each gearbox.<\/li>\n<\/ol>\n<\/li>\n<\/ol>\n

          This combined approach limits the need for physical prototypes, reduces development time, and improves the quality of the finalized product.<\/p>\n

          Want to know more? Join us next week at the 11th Annual Offshore Wind Europe Conference & Exhibition<\/a>\u00a0in London, UK. Wim Hendricx, Simcenter Engineering expert for the Energy sector, will present this application case on November 28 at 9:20 AM.<\/p>\n

          \"Wim-Hendricx-Winergy-quote.jpg\"<\/span><\/a><\/p>\n

          Interesting links:<\/p>\n