Question on the selection of threshold in "Wind Turbine High-Speed Bearing Prognosis" example.

Question on the selection of threshold in "Wind Turbine High-Speed Bearing Prognosis" example.

If the historical data is available, for example, there are 100 wind turbines with run-to-failure historical data. Is there any way to calculate a "common" threshold for the 100 wind turbines as well as the new coming turbines instead of selecting last day as the threshold? Since each wind turbines may have different'last day'.

The predictive maintenance examples and predictive maintenance toolbox are very impressive and really practical , thanks a lot.

Miracle

[Image Analyst]

Yes it is.

If you have any other questions, just read this link then ask your next question.

For reference, I'll post the link you forgot to post: https://www.mathworks.com/examples/predictive-maintenance/mw/predmaint-ex34579581-wind-turbine-high-speed-bearing-prognosis?s_tid=answers_rc2-2_p5_MLT

[Pengcheng Fang]

Hi Image Analyst，

Thanks for your reply and your help of posting the example link.

Could you provide more detailed information about how to calculate the threshold ?

Since it seems the way of calculating threshold based on historical data is not introduced in this example.

Thanks

Answer 1

[Xiaoxing Wang]

If you have 100 samples, instead of calculating the threshold, you can try the similarity model to estimate RUL as an alternative approach.

https://www.mathworks.com/help/predmaint/ug/similarity-based-remaining-useful-life-estimation.html