#1 2017-02-22 15:03:40

AleDela
Member

Weibull VS Exponential Law

Good afternoon,

First of all, thank you for your software and your service:)

I did a test:

- Using the exponetial law for all the base events
- Using the weibull law for all the base events applying Beta=1 and alpha=MTTF=1/lamda (case: Weibull= Exponetial)

But I do not get the same results (different F(t)). I performed a "calcul ponctuel".
Could you help me to understand where is my error.

Thank you for your time

Best

Alessio

Offline

#2 2017-02-23 09:57:57

Emmanuel
Administrator

Re: Weibull VS Exponential Law

Good morning Alessio,

I did a simple test and everything seems to be ok for me.

Hereafter, you will see a screenshot.

436425Capture.png

The Weibull law has the same result than the exponential law.

Are you ok with this test ?

Emmanuel

Offline

#3 2017-02-23 11:16:28

AleDela
Member

Re: Weibull VS Exponential Law

Good morning Emmanuel,

Thank you for your answer. Unfortunately I tried again this morning but I can not obtain the same results.
I think that I am doing the same approach, using more gates and base events.
However, I am attaching my FTA...

test-law-exp-vs-weibull_imagesia-com_1d793_small.png
Hébergé par Imagesia.com

What do you think?

Moreover, I would like to understand why the plot does not follow the classic exponetial law trend and I do not obtain a costant lamda during time ( lamda=1/MMTF if I use the exponetial law).

I think that I am making some errors but I do not know where exactly.

Thank you for your time

Best

Alessio

Offline

#4 2017-02-23 11:55:50

Emmanuel
Administrator

Re: Weibull VS Exponential Law

Alessio,

1/5e-6 => 2e5
1/2e-6 => 5e5
1/3e-7 => 3,33e6

After theses modifications, you will find the correct result. smile

Offline

#5 2017-02-23 13:46:28

AleDela
Member

Re: Weibull VS Exponential Law

Emmanuel,

Thank you so much, I made a big mistake sorry:)

Can I have a suggestion?

Concerning Weibull law, If I would like to consider the fatigue phenomena I have to use beta>1. However, I am not sure which value I could apply to the scale parameter alpha. I know this relationship: MTTF= alpha Γ (1+1/beta).
Knowing that I have the MTTF of each base event are there any other way to get the parameter alpha? Or I have to use this formula: MTTF= alpha Γ (1+1/beta).

Ps: why if I always use for each base event an exponetial law the global plot does not follow the classic exponetial law trend and I do not obtain a costant lamda system during time ( lamda=1/MTTF).

Sorry for all these questions and thank you so much for your time and kindness

Best

Alessio

Last edited by AleDela (2017-02-24 09:30:24)

Offline

#6 2017-02-24 11:15:42

Emmanuel
Administrator

Re: Weibull VS Exponential Law

I think I'm not the right person to make a suggestion about the fatigue phenomena.

However, the global plot depends on the structure of your fault tree (OR gates, AND gates etc...). It's normal that the top gate's curve differs. You can easily reproduce that on Excel using a simple AND gate.

About a constant system lambda, you can read this thread http://www.arbre-analyste.fr/forum/viewtopic.php?id=32.

Regards.

Offline