I have got a likely pretty straightforward question.
Let assume that I have a DEL value weighted over 20 years and a second value weighted over 0.25 years: is it possible to get a overall DEL weighted over a 20.25 years lifetime, starting from the two above? Nref = 1E+08 cycles and Woehler slope m=4.
The damage-equivalent load (DEL) is defined as follows:
DEL = ( SUM( n_k*L_k^m, k = 1, 2, …, N )/n_eq )^(1/m)
where,
k = load-range bin counter
N = number of load-range bins
n_k = number of cycle counts in the k’th load-range bin
L_k = k’th load range
n_eq = number of equivalent cycle counts
m = Whoeler material exponent
I’m not exactly sure how you are defining Nref. I’m assuming you mean that Nref is the number of equivalent cycle counts per year, such that 20*Nref is the number of equivalent cycle counts in 20 years.
I’m also not clear on how you are defining DEL(20) and DEL(0.25). If you mean that DEL(20) is a DEL based on Nref that acts for 20 years out of a 20.25-year period and DEL(0.25) is a DEL based on Nref that acts for 0.25 years out of a 20.25-year period, and my assumption about Nref is correct, than DEL(20.25) is:
your thoughts about Nref, DEL(20) and DEL(20.25) are absolutely right. Indeed, I myself came up with the same formula you wrote down 10’ after posting my question.
I just wanted to have my idea supported by somebody who has a more consistent knowledge than me.
Can I estimate the Lifetime DEL based on Short-term DEL using the following method?
Annualized DEL = Sum(DEL(i) * f(i)) for i = 1, …, n,
where DEL(i) is the ith Short-term DEL, and f(i) is its corresponding frequency.
Lifetime DEL = Annualized DEL * Tlife
where Tlife is the design life of the equipment. @Jason.Jonkman
MLife can use the Weibull wind distribution; it can also use a user-specified distribution, including multiple dimensions. See the documentation User_Probability_Distributions.pdf in the MLife archive for more information.
I’m still not sure I understand your equation for lifetime DEL in your step 3, which seems to differ from what MLife uses, but I agree with steps 1 and 2 in your method. Also, 20 seeds in step 1 may be more than you need (the IEC design standards only require 6 seeds, although it is questionable whether 6 is enough to ensure statistical convergence; 10 seeds may be more reasonable).
Dear Jason,
I have read the User_Probability_Distributions.pdf file and know the file format of distribution table, but how should I go about generating the binary file of distribution table?
In the MLife archive, there is a MATLAB function write_joint_pdfm.m that can be used to write the user-specified distribution to a binary file used via DistribName in MLife.
Is my below understanding of the number of equivalent cycles correct?
The n_eq value is a value that we predetermine and then keep consistant throughout or DEL calculations for each load history of interest, and we keep that consistant to make load histories of different cycle numbers comparable.
With this in mind, would you consider this python library for DEL calculation to be adequete. The rainflow counting method seems to check out, so I guess it is good?