CWD
  • Conference for Wind Power Drives 2017
    Conference for Wind Power Drives 2017
  • Aufbau FVA-Gondel
    Aufbau FVA-Gondel
  • Conference for Wind Power Drives 2017
    Conference for Wind Power Drives 2017
  • Center for Wind power drives
    Center for Wind power drives
  • Conference for Wind Power Drives 2017
    Conference for Wind Power Drives 2017
  • Aufbau FVA-Gondel
    Aufbau FVA-Gondel
  • Vorstellung des 4MW Prüfstandes
    Vorstellung des 4MW Prüfstandes
  • 4MW-Prüfstand mit HybridDrive
    4MW-Prüfstand mit HybridDrive
  • 1MW-Prüfstand
    1MW-Prüfstand
  • Campus Melaten
    Campus Melaten
23.05.2019 16:51
Category: Allgemeine News

Risk Evaluation of Lubricant Compounds Causing White-Etch Cracking Using Artificial Neural Networks


Artificial neural networks (ANN) help identfying high risk lubricant compouds.

Thinking the future of wind power: our colleague Baher Azzam will be presenting on risk evaluation of lubricant compounds using artificial neural networks at the International Conference on Gears 2019. You can find out more about this exciting topic already today: Artificial Intelligence (AI) is being applied in various industries while its self-learning ability allows users to optimize processes. The characteristics of 700 different oils had been provided to the CWD by 4LinesFusion in order to be analyzed as part of project ProNOWIS. This study’s goal consisted in using state-of-the-art artificial neural networks (ANN) models to assess the output from 4LineFusion’s SeerWorks™ Reliability database application.  At first, the composition of the 700 oils was processed in SeerWorks™ Reliability to provide output data on the type of white-etching cracks (WEC) class (low/high risk); the type of WEC class is associated with a certain combination of additives that comprises the lubricants in relation to their interaction with the surface of the bearing material. In the course of the study, the lubricant additive compounds were blinded and labelled by identification numbers to provide neutral dimensionless values for ANN processing. Firstly, the ANN processing was to evaluate the White-Etching Crack (WEC) risk classification levels based on the oil identity and percentage of constituting compounds. Secondly, oil compounds that affect the WEC risk in gearbox bearings were identified. The key outcome of this analysis has been the ability of ANN models to identify 8 out of 21 oil compounds as highly influential on WEC risk. Several ANN models were developed, gradually increasing the classification accuracy on test oils to 99.8% by altering the network’s architecture. Reliability can provide essential information on WEC risk in an application to prevent upcoming WEC failure before any damage has occurred, and thus has a profound effect on operational and maintenance costs as well as the availability of wind turbines. Find out more about the publicly funded project ProNOWIS.