RAINFLOW COUNTING METHOD FOR GNU OCTAVE

Introduction

Description

An alert for the engineer calculating fatigue damage

• the stress variation, the amplitude
• the average stress value for the cycle
• the number of cycles (0.5 or 1)

Observe that the stress range, the double amplitude, is usually applied in SN-curves when calculating the fatigue damage by the Miner-Palmgren method. You should normally do: range = 2*amplitude

Rainflow algorithm

The algorithm used is the "range pair-range" algorithm as described in ASTM E 1049-85 and by de Jonge (1982). The result from the "range pair-range" algorithm is exactly similar to that of the pagoda-roof "rainflow" algorithm as described by Matsuiski, M. and Endo, T. (1969). However, their descriptions are very different and the "range pair-range" algorithm is very much simpler to program.

References

1. Matsuiski, M. and Endo, T. (1969): "Fatigue of metals subjected to varying stress"
2. de Jonge, J.B. (1982): "The analysis of load-time histories by means of counting methods"
3. ASTM E 1049-85 (Reapproved 1997): "Standard practices for cycle counting in fatigue analysis", in: Annual Book of ASTM Standards, Vol. 03.01, Philadelphia 1999, pp. 710-718.
4. Wikipedia: Rainflow-counting algorithm

Author

Package for GNU Octave

Figures

The following figures show an example rainflow matrix and stress cycles extracted from a series of turning points.

Compilation and installation for GNU Octave

1. Do the following to create a compiled mex-file from "rainflow.c" and to test the result:
   $ cd ./rainflow-octave
$ make
$ ls ./src/rainflow.mex
rainflow.mex
$ make test

2. Move files: rainflow.mex, rainflow.m and sig2ext.m to a place where GNU Octave can find them.
   Octave will locate oct- or mex- files automatically if they are in a directory listed in the search path. To figure out the search path, do:
   $ octave
octave> path
Octave's search path contain the following directories:
   .
   /usr/local/share/octave/site-m
   

Disclaimer for GNU Octave

1. The rainflow function.
2. The function for extracting turning points from stress time series.
3. Documentation and testing.
4. Packaging.

The auxilliary demonstration scripts are fully unmodified and have not been tested extensively. The script rfdemo1.m is suspected to contain a Matlab specific command associated with a button. The GNU Octave version 3.0.0 has been applied.

Enjoy!
Torfinn Ottesen

The below text was written by Adam Niesłony before the GNU Octave packaging. It is kept unmodified for reference purpose. For usage instructions, see the below Quick start.

Rain flow for MATLAB

Version date: 12.01.2005

Description

Here you can find the rain flow algorithm for cycle counting, adapted for work in the MATLAB^® environment. It has been translated from Turbo Pascal into C language and compiled to the MEX function. The algorithm code has been prepared according to the ASTM standard [1] and optimized considering the calculation time. The function was tested with use of PC computers in the MS Windows^® operating system.
Please send all your remarks, suggestions and questions to the author.
E-mail: Adam Niesłony

Author

The MEX function of the rain flow algorithm has been elaborated by Adam Niesłony. At present he is a postgraduate student of Technical University of Opole and he works on his Doctor's thesis in Department of Mechanics and Machine Design.
See the following pages:
Technical University of Opole
Katedra MiPKM

Files

Toolbox files:

rainflow.m   - short description of the rainflow function,
rainflow.c   - C MEX code of the rainflow function,
rainflow.dll - compiled "rainflow.c" file (MATLAB 6.5, Windows 2000 Prof.),
rfhist.m     - histograms of rainflow data,
rfmatrix.m   - rainflow matrix calculation and visualization,
rfdemo1.m    - demo for cycle counting, recomended for short signals,
rfdemo2.m    - demo for rainflow matrix and histograms, recomended for long signals,
sig2ext.m    - auxiliary function for rfdemo, it searches turning points from signals.

Installation

In order to instal the rainflow function, you must realise standard tasks, necessary during addition of new toolboxes to the MATLAB, i.e.

1. Download the file rainflow.zip
2. Create a new directory for files of the rainflow toolbox
3. Unpack all files from the archive to that directory
4. Add a path to that directory, using the command "Edit Path" of the MATLAB

Quick start

Using demos:

clear all
rfdemo1
rfdemo2

Rainflow counting on self generated (or loaded) signal:

clear all
lengths=10000;
s=10*randn(lengths,1)+rand(lengths,1);
% or load the signal from file --> s=my_signal;
tp=sig2ext(s);
rf=rainflow(tp);
rfm=rfmatrix(rf,20,20);
surf(rfm)

References

end of page
© by Adam Niesłony