MACHINE LEARNING

MACHINE LEARNING APPLICATION

 Gas turbines are complex, crucial and critical machines. Protecting, optimising and advancing the operation of these machines is important to our customers and operators of these engines.

By utilising engine data (already available) we can achieve the following:

  • Data acquisition via existing controller
  • Data fed to machine learning model
  • Machine learning model determines various conditions based on data

This data being made available to a mathematics model will:

  • Increase life of components
  • Reduce engine downtime
  • Notify user of conditions which will prevent catastrophic failure
  • Extend interval between overhauls
  • Significantly reduce opex budgets

The mathematics model utilises combinations of the following techniques:

  • Singular spectrum analysis.
  • Fast Fourier transform.
  • Wavelets.
  • System entropy.
  • Bayesian probability.

to determine the following:

  • Deviations from safe operating conditions.
  • Deviations from optimal operating .
  • Digital twin capabilities.

These conditions are mitigated through constant monitoring and analysis, allowing existing data streams to be used more effectively in condition based analysis.