Dear Fahime Sokhangou

I have worked with ARTeMIS for SHM purposes, do not hesitate to contact me and ask your questions in field of OMA, EMA and so on.

Good luck.

Hossein Safar Yousefifard Thanks for replying me. In fact, I did experimental modal analysis in Artemis for the series of measurements that I have done for a damaged structure ( The method is roving hammer and I had 2 reference accelerometers).

the mode shape I obtained is not smooth and I wonder if there is measurement noise.

How I can improve and decrease the noise?

Is the position of the reference accelerometer matter in the roving hammer method?

Your welcome.

About your first question, there are many methods to reduce noise in a signal but one the powerful one is high order butterworth filter which you can use it to filter some frequencies but you should be careful when you utilize it because you might eliminate some points of actual signal.

About your second question, sensor placement and array has been a challenging issue in SHM, but in your study it is not obvious the effects of sensor array i suggest you if it is possible, change the sensor placement and then compare them to see whether it has high effects on your mode shape results or not.

Of course in signal processing there are some methods for nonlinear and non-stationary signals which decompose the signal to its components and then they reconstruct it which leads to the more linear and more stationary signal. for example empirical mode decomposition (EMD) and its variants, variational mode decomposition (VMD). you can test the effects of them in your study specially VMD.

Hossein Safar Yousefifard Thanks,

In my project, I have 12 accelerometer is different positions and I selected 2 of them as the reference).

I can select the references from the other sensors, but about the signal processing, How I can perform it in Artemis?

Should I reduce the noise before importing the data in Artemis?

if yes, how?

12 accelerometer is enough for alleviate uncertainty but you can consider other sensors as reference to see the changes.

All of the above mentioned signal processing methods should do before importing to ARTeMIS. At first reduce the noise using butterworth filter in MATLAB and extract the filtered signal afterward apply EMA using ARTeMIS on filtered signal.

you can search on google, github, stack overflow and other websites to find MATLAB and Python codes and scripts for noise reduction methods,

Hossein Safar Yousefifard Thank you very much. I have been considering the possibility of performing signal denoising in Artemis. Are you suggesting applying noise reduction techniques to the time-domain data obtained from accelerometers? If so, how can we ensure that these noise reduction methods do not eliminate the effect of damage?

another question, for a simply supported beam, the second mode shape is like sin(x). I exported the second mode in abaqus but it gave me the ABS value.

Do you have any idea how I can export real numbers (with sign)?

sure. Yes you should perform denoising method in time domain and then start to work on your acceleration signals.

one way to see that you have eliminate the noise ( not main signal point )is to observe modal frequency of structure before and after the damage occurrence. you should be careful that do not eliminate frequencies close to your structure frequency at different modes.

unfortunately I haven't worked with Abaqus software and I can not suggest you any idea. but if you can it is better to simulate your simple beam in MATLAB which is very popular and you can use scripts in github or other websites to perform all of your analysis and get your results.

Hossein Safar Yousefifard Thank you so much, but my data is real measurements and does not come from Abaqus.

I did the modal analysis and I know the resonance frequencies.

Hossein Safar Yousefifard Dear Hossein, if you see the attached image you will see that in signal processing section in Artemis, there is some filters but if I enable them, the run bottom does not work anymore.

Reduce the upper cut-off frequency of filtering, see if the problem is solved.

Unfortuanately I haven't faced with such problem but chage some options in this window to see whether the bottom will activate or not.

If are still face with running problem you can filter your signals in MATLAB the below link will help you to do that in MATLAB:

https://www.mathworks.com/help/signal/ref/butter.html

Mohammad Sadegh Ayubirad Thank you so much. it worked.

Hossein Safar Yousefifard thank you so much. I will do denoising first. But I am afraid if I remove some information about the damage.

I would like to use the first mode shape for damage detection algorithm that I have developed.