Dear researchers,
Experimental investigations are to be carried out on a prototype to determine the design stress levels.
Prototype : Chamber filled with SF6 gas. Conductor running longitudinally and isolated from the tank/chamber by the gas.
Objective: Determine the withstand voltage level with 50% probability through experimentation and establish the 90% withstand voltage level theoretically.
Reason : In up-down method of stressing the insulation, only 1 shot per group is to be applied and there is no mention in the IEC 60060-1 to determine the 90% withstand voltage level.
We know that the breakdown characteristics of SF6 follows weibull distribution. Could I determine the 50% withstand voltage level and extrapolate through curve fitting to obtain 90% withstand voltage level?
If yes, how do I proceed after establishing 50% withstand voltage level? How to determine the maximum likelihood estimator and how is it helpful in this regard?
Thank you
I would proceed in this way: after you have done the up-and-down test you have a number of breakdown voltages. From these values you can determine a weibull distribution and gain the 90% withstand voltage.
There is an easy tool for fitting the weibull distribution avaliable: Visual-XSel where there is a free version for download under www.weibull.de
Thank you for the response. With what confidence level do I determine the U50? What is the confidence attached to the U90 that has been obtained from the wiebull distribution? Thank you for the tool.
Of course, you can estimate the Weibull percentiles with at a given confidence level. We published a paper on how to estimate these percentiles for the 2-parameter Weibull distribution. It is available for download at ResearchGate.
Thank you for providing the paper. This would help a lot of people like me. Good job sir.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics