Dear sir! help me !About the threshold!!!

Everything about quantification of recurrence plots and recurrence networks.
Post Reply
Yuzhi233
Junior
Posts: 3
Joined: Wed Mar 10, 2021 04:21
Affiliation (Univ., Inst., Dept.): zhengzhou university Henan China
Location: china Henan zhengzhou
Research field: Bearing fault diagnosis

Dear sir! help me !About the threshold!!!

Post by Yuzhi233 »

Dear sir:
I am doing some research on bearing fault diagnosis by using RQA ,like the guy in the forum ,I am very confused about the selection of the threshold epsilion.
so I use CWRU bearing fault data ,calculate the Normal bearing and faulty bearing's RPs,but I can't determine the threshold (i’ve seen your recommended papers ),I am trying a lot of threshold but I do not know what is the best choice (sequence's std 0.25*std 1*std,1.5*std,2*std,even 10% max phase radius).
if i want to to compare the different RQA feature between the different bearing fault,how should I set the threshold? Can I simply set them as a factor multiple of the standard deviation of the time series,or 10% max phase raduius?(I found normal status‘ rps use the 10% max Phase radius almost white and one of the fault condition is almost black..)
i am very sorry about my stupid and repeat question ,i dont know how to set the threshold to makes the recurrence structure clear ...
User avatar
Norbert
Expert
Posts: 194
Joined: Wed Jan 4, 2006 11:03
Affiliation (Univ., Inst., Dept.): Potsdam Institute for Climate Impact Research, Germany
Location: Potsdam, Germany
Location: Potsdam Institute for Climate Impact Research, Germany

Re: Dear sir! help me !About the threshold!!!

Post by Norbert »

The threshold selection depends on the research question and the data.

If your data has many outliers or behaves like from a non-stationary process, a normalisation and using a threshold in units of the standard deviation might not be the best choice. Then consider the slecetion that preserves a preselected recurrence rate (such as based on a quantile of the pairwise distances or the fixed amount of nearest neighbours approach).
Post Reply