How can i do frequency weighted to acceleration time domain data to assessment the whole body vibration as (ISO 2631-1) recommended?

89 visualizaciones (últimos 30 días)
Hi i need matlab code for frequency weighted acceleration as ISO 2631-1 plz.
Attached: sample data sheet (x, y, z axes m/s^2), sample rate 0.005 sec.
  3 comentarios
Yahia Gaddal
Yahia Gaddal el 15 de Ag. de 2021
This is a first time for me to use matlab. My study about assessment of whole body vibration, i have vibration data from field as acceleration (m/s2) the next step should be convert this data from time domain to frequency domain and weighting the frequency as ISO 2631-1:1997(E) recomended for whole body vibration assessment. So i am trying to use Tomirvine999 "vibrationdata" function in Matlab GUI package, but i get this warrning (nnn>50000) when press on calculater botton.
William Rose
William Rose el 16 de Ag. de 2021
The help you are requesting requires a lot more than the usual amount of effort for answers on this site.
I downloaded the vibrationdata package (1_1_2021 vrsion) from Tom Irvine's site. When I unzipped it it was over 100 MB. I ran vibrationdata in Matlab. The GUI opens. You said "i get this warrning (nnn>50000) when press on calculater botton". However, I do not see a calculator button. I see the screen below.
I assume you loaded some of your own data, and then selected various options, or clicked on some the buttons and menus on the screen. But I don;t know what you did, and I do not see a calcualtor button.
I think that if you have quesitons about this package, you should ask its author, Mr. Tom Irvine.
I tried the link you posted for ISO 2631-1. The link goes to a long list of ISO standards, one of which is ISO 2631-1:1997. I clicked on that link. I would need to pay 138 CHF to see the standard. If you have access to the standard, I would check the following: Does the standard recommend a specific sampling rate, or a range for sampling rate? Does the standard recommend a specific sample duration, or a range for sampling duration? Does the standard recommend adding up all the powers in thespectrum over a certain range of frequencies, with different wieghts for different frequencies? If so, what is the range? What are the weights? Acceleration is a vector. How does the standard reocmmend dealing with the vectorial nature of acceleration? For example, does the standard recommend measuring x, y, and z components of acceleration separately, and analyzing each separately? Or does the standard recommend measuring the instantaneous magnitude, and ignoring the direction? Or some other approach?
Do you have a file of time-domain vibration data that you can post? If so, please epxlain the organization of the file: identity and units of measure for each column, and sampling rate, if there is not a time column.

Iniciar sesión para comentar.

Respuesta aceptada

William Rose
William Rose el 16 de Ag. de 2021
Thank you to @Simon Chan for reading your quesiotn carefully and for noticing that you had posted a data file and that you have posted the smapling rate. Sinc you are new to Matlab, I will describe some steps to import and view your data and prepare it for analysis in the 'vibrationdata' package.
accelData=importdata('acceleration.csv');
disp(accelData);
data: [1155×3 double]
textdata: {'X' 'Y' 'Z'}
colheaders: {'X' 'Y' 'Z'}
Since acceleration.csv includes text and data, Matlab imports it into a "structure", which can contain more than one type of data. I used the "disp" command to see what is inside "accelData". I learned from the reuslts of the disp command that there is a 1155x3 array of double precision numbers inside accelData, and the array name is accelData.data.
dt=0.005; %time step size
t=0:dt:1154*dt; %create a vector of times
accelX=accelData.data(:,1); %X component of acceleration=column 1 of array accelData.data
accelY=accelData.data(:,2); %Y component=column 2 of accelData.data
accelZ=accelData.data(:,3); %Z component=column 3
accelR=sqrt(accelX.^2+accelY.^2+accelZ.^2); %compute resultant acceleration
figure; %open a blank figure window
subplot(2,1,1); %divide figure into 2 rows by 1 column; prepare to plot in top panel
plot(t,accelX,'r',t,accelY,'g',t,accelZ,'b'); %plot, X, Y, Z versus time
grid on; legend('X','Y','Z'); %add grid and legend
ylabel('Accel (m/s^2)'); %add axis label
subplot(2,1,2); %divide figure into 2 rows by 1 column; prepare to plot in bottom panel
plot(t,accelR,'k'); %plot resultant acceleration versus time
grid on; legend('Resultant'); %add grid and legend
xlabel('Time (s)'); ylabel('Accel (m/s^2)'); %add axis labels
The commands above generate the figure below.
I learned yesterday that 'vibrationdata' accepts an array that has a time column and a data column. Therfore I create two such arrays, for X acceleration and resultant acceleration:
dataX=[t' accelX]; %transpose t, to make it a column vector, to match accelX
dataR=[t' accelR];
Then I run 'vibrationdata'
vibrationdata
I make the selections shown below and click "Begin Signal Analysis".
In the next pop-up window, I select "ISO 2631 Human Vibration", and I click "Analyze".
In the next pop-up window, I select Units=m/s^2; Mean Removal=Yes; Weight=Wk, Z-axis, Seat; Subdivide into Segments=No. I specify input array name=data1, start time=0, end time=6. I click "Calculate". This produces a plot like the one shown by @Simon Chan:
You can repeat for the resultant aceleration, and so on. As Mr. Chan correctly notes, the 'vibrationdata' package has many options.

Más respuestas (2)

William Rose
William Rose el 16 de Ag. de 2021
Editada: William Rose el 16 de Ag. de 2021
I decided to test the vibrationdata package of Mr. Tom Irvine by creating some simulated acceleration data. My data is random noise with a white (i.e. flat) spectrum frequency range from 1 to 20 Hz. The simulated data is recorded at 160 Hz for 1 minute. Here is a script to make the simulated data with the specified properties:
clear;
fs=160; %sampling rate (Hz)
Td=60; %duration (s)
N=Td*fs; %number of samples
dt=1/fs; %sampling interval
df=1/Td; %frequency spacing of the signal, in the frequency domain
t=0:dt:Td-dt; %vector of times
f=0:df:fs-df; %vector of frequencies for the discrete Fourier transform
Xf=zeros(1,N); %initialize array Xf(), for the signal's discrete Fourier transform
%next line makes the spectrum have magnitude 1, and random phase, from 1 to 20 Hz
Xf(f>=1 & f<=20)=ones(1,length(f(f>=1 & f<=20))).*exp(1i*2*pi*rand(1,length(f(f>=1 & f<=20))));
%Next line makes the spectrum from 140 to 159 Hz be the complex conjugate of the spectrum from 20 to 1 Hz.
%This is required by the mathematics of the discrete Fourier transform, in order to obtain a time domain signal that is real and not complex.
Xf(f>=140 & f<=159)=conj(flip(Xf(f>=1 & f<=20)));
Plot the magnitude and phase of the signal's DFT to confirm that it has the desired properties: constant magnitude, and random phase angles, from 1 to 20 Hz, and Xf(f=140 to 159 Hz) is complex conjugate of X(f=20 to 1 Hz).
figure;
subplot(2,1,1); plot(f, abs(Xf),'.r'); ylabel('Magnitude');
subplot(2,1,2); plot(f, angle(Xf),'.r'); ylabel('Phase Angle (radians)'); xlabel('Frequency (Hz)');
Compute the time domain signal by inverse FFT:
xt=ifft(Xf);
Plot the entire time domain signal, and the first and last 1 seconds:
figure;
subplot(2,1,1); plot(t,xt,'.b'); xlabel('Time (s)'); ylabel('Amplitude');
subplot(2,2,3); plot(t(t<1),xt(t<1),'.-b'); xlabel('Time (s)');
subplot(2,2,4); plot(t(t>=59),xt(t>=59),'.-b'); xlabel('Time (s)');
The time domain signal xt looks reasonable. It is real and randm-looking and there are not fluctuaitons with frequencies that are obviousy higher than 20 Hz. Save it in a text file so we can read it in to Tom Irvine's 'vibrationdata' package.
save('simVibData1.txt','xt','-ascii');

William Rose
William Rose el 16 de Ag. de 2021
I tried using the file generated by the code above in'vibriton data'. After clicking sevral buttons I reached a screen that says the data file should have two columns: time and acceleration. And after experimenting, I found that it 'vibraitondata' looks for an array in memory, rather than a file on disk. Therefore I am changing the last line of hte code I posted above, from
save('simVibData1.txt','xt','-ascii');
to
data1=[t',xt'];
Then I make selections in 'vibrationdata': Under "Signal Analysis Functions", I select "Time History" and "Acceleration" and "ISO Generic VC, 2631, 10816'. Then I click the button "Begin Signal Analysis". A new window pops up. In that window, I select "ISO 2631 Human Vibration". Then I click "Analyze". A new window pops up. In that window, I enter Input Array Name=data1 (because I have alrady created that array with my script), Unit=G, Mean Removal=Yes, start time=0 sec, end time=60 sec, Weight='Wk, Z-axis, seat', Segment Duration=10 sec, Percent Overlap=50. Then I click the button "Calculate". Three figures pop up see below, and a window pops up that says "Results written to command window.". I click the "OK" button in that window and it disappears.
I had the same result as you ("Warning: nnn>500000") when I left start time, end time, and duation fields blank, and then I clicked "Calculate".
  3 comentarios
Simon Chan
Simon Chan el 16 de Ag. de 2021
Editada: Simon Chan el 16 de Ag. de 2021
As mentioned in the description, the sampling rate is 0.005 sec. Hence a time stamp can be created based on the number of samples (in this case, it is 1155). So the sampling time starts from 0 sec to 5.77 sec).
Then, separate input data are created, for example, Accx, Accy & Accz which consists of one column of time stamp and one column of acceleration data.
Execute the given GUI and choose [Wd, X or Y-axis, Seat] for both Accx and Accy, I got the following:
(I have no idea about ISO 2631-1 and not sure the following results are correct or not, so just sharing)
Then selecting [Wk, Z-axis, Seat] for Accz, I got the following:
However, there are so many options in the GUI and hence it really depends what the user want to have.
William Rose
William Rose el 16 de Ag. de 2021
The y axis label for the three plots posted by Mr. Chan is "Accel (m/sec^2) RMS". But the plots show negative values. An RMS (Root of the Mean of the Squared) value should never be negative. Therefore the values on these plots cannot really be RMS values. If I were using the 'vibrationdata' package, I would investigate this further, by analyzing my own data in Excel versus in 'vibraitondata', and by generating simulated data with known properties, and analyzing it in 'vibrationdata', and so on.

Iniciar sesión para comentar.

Categorías

Más información sobre Vibration Analysis en Help Center y File Exchange.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by