rainflow

[MvdH]

Hi,

I'm fighting with the rainflow procedures in Diadem 10.2.

I would like to get two colums for each channel, the first one with the range (class) and the second column with the number of counts. I prefer this over the matrix setup which is often used, but of course I can build these two channels once I have the matrix. The problem is that the output seems to be incorrect.

I have two example data sets with example output of how the rainflow matrix should look, but when I try this I do not get the expected answer.

1st example:

FVA Merkblatt 0/14 dataset: 8,3,4,1,8,7,8,2,8,5,7,1,8,2,4,3,8

Following the explanation from FVA this should bring 2 counts in class 7, 2 counts in class 6, 1 count in class 2 and 3 counts in class 1 following the method by Matsuishi. The example matrix given in FVA shows the same result and since Diadem is supposed to be based on FVA it should of course give the same result with a total of 8 counts. Note for the matrix below: 'from' class on vertical, 'to' class on horizontal lines.

   1  2  3  4  5  6  7  8

1  0  0  0  0  0  0  0  0

2  0  0  0  0  0  0  0  0

3  0  0  0  1  0  0  0  0

4  0  0  1  0  0  0  0  0

5  0  0  0  0  0  0  1  0

6  0  0  0  0  0  0  0  0

7  0  0  0  0  0  0  0  1

8  2  2  0  0  0  0  0  0

When I run this set through Diadem I get a total of 8 counts again, but a different matrix. Fyi classes are set to automatic with no. of classes set to 8 and closing and inclusion of residues is checked. Note for the matrix below: 'X' on horizontal, 'Y' on vertical.

   1  2  3  4  5  6  7  8

1  0  0  0  0  0  0  0  0

2  0  0  0  0  0  0  0  0

3  0  0  0  1  0  0  0  0

4  0  0  1  0  0  0  0  0

5  0  0  0  0  0  0  1  0

6  1  0  0  0  0  0  0  0

7  0  0  0  0  0  0  0  0

8  1  2  0  0  0  0  1  0

This equals 1 count in class 7, 2 counts in class 6, 1 count in class 5, 1 count in class 2 and 3 counts in class 1 which is a very significant difference if I were to use it for fatigue calculations.

Of course I have tried mulitple different approaches, but the result is never as expected. Especially interesting is that when I change to manual class setup the whole matrix shifts which makes me wonder what I'm missing here.

I have also tried the example set which is in this board (3,1,4,3,4,2,4,3,5,1,2,1,4,2,3,2,4). Needless to say that the result is again not as suggested, so please help. Should any of you want to read up on the answer to that dataset just search for "FVP Rainflow Documentation" on this board. If you can replicate the answer given there, please advise which settings to use since I can't find them.

Thanks,

Marcel

[Christian_S]

Hi Marcel,

DIAdem gives a different result as you expect because - by default - the first reversal point which crosses a class limit is counted as the first valid value in the load time function. The last value of the load-time function is counted as invalid.

Since DIAdem 2010 (11.2) there is an new switch RainInclFirstLastVal which specifies whether DIAdem uses the first and the last value of the signal as the reversal points in a rainflow classification.

I have attached an example which at first calculates a transition matrix (not needed here) and a rainflow matrix from your example load-time function (FVA Merkblatt 0/14 dataset) with the build-in DIAdem function ChnRainCalc. This piece of code is the same you get when running the rainflow calculation interactively with teach-in mode enabled.

Pay attention that there is the new parameter RainInclFirstLastVal used which is introduced with DIAdem 2010.

In a second step the range-mean matrix is calculated from the rainflow matrix with a small script function.

The load-time function and all results are shown on several pages of a report layout.

I hope this will help. 

Regards,

Christian

[Christian_S]

Sorry, I forgot the attachment...

[MvdH]

Hello Christian,

Thank you for the reply on this one. If I understand correctly Diadem will not give the answer I expect unless we upgrade to the 11.2 version in which some additional features are available. Since Diadem help mentions it uses the FVA 0/14 procedure I would have expected differently, but maybe one of the developers can answer that one.

In the meantime I think you are probably using a higher version than my 10.2 since the script will not run. It has some problems with the object structures, but I may be able to work around that, so thanks so far and hopefully I can report that it works after this weekend.

Thanks,
Marcel