Lower cutoff frequency for signal processing (filtering ) of LVDT and strain gauge data?

4 views (last 30 days)
PIYUSH MOHANTY on 19 Sep 2019
Commented: PIYUSH MOHANTY on 24 Sep 2019
I have carried out some shake table tests on piles inside the soil. I have applied white noise motion as input for the shake table. While estimating the bending moment and displacement, what should be the lower cutoff frequency while applying filter?
I usually do a low pass filter for LVDT data and strain gauge data, where i define the upper cutoff frequency. I do not put a lower cutoff frequency as putting a lower cutoff frequency will take out the bending moment or displacement of a pile, which may be of residual value or basically static in nature.
Hope you understand this point. Is my approach correct?
PIYUSH MOHANTY on 19 Sep 2019
Thanks thoughtGarden, I completely agree that a low pass filter will definitely work, as it does not take away the residual displacement value.
Some of my peers advise me to take a bandpass filter with a lower cutoff frequcny of say 0.5 Hz. But this approach will take away the residual values (which are static in nature). That's why I came up with this doubt.
I am having strain gauges, from where I can straight away calculate the bending moment.
By the way how do we estimate the right upper cutoff frequency for a low pass filter? Any approach/ideas...

Sign in to comment.

Answers (1)

thoughtGarden on 19 Sep 2019
It sounds like a low pass filter is all that is needed. There are several considerations in selecting low-pass cutoff frequencies. Given that you know something about the input (shaker table motion) and the structure, you can select the filter type and bandwidth.
Considering that I don't know anything about your setup, I will quote from Mechanical Vibrations and Shock Measurments, "It is not possible to give generally valid rules for selection of bandwidth".
That being said, here are two potentially invalid rules to consider
  1. Don't let your filter remove the real thing...the mechanical reasonance of the instrumented structure. You want to capture the strain/deflection of the structure, so the bandwidth must be greater than the -3dB frequency of the structure. If possible, you want seperate the filter and natural frequncy by as much as possible while still achieving the goal of removing noise. This is a balancing act which is custom to each application.
  2. If digitally filtering, always be mindful of aliasing. Aliasing is when high frequncy content appears as low frequency folded/ghost content due to sampling frequency being below 2*signalFrequency.
There are other guidlines, but start here and return with more questions.
PIYUSH MOHANTY on 24 Sep 2019
Hi ThoughtGarden Thanks for your reply. I completely get your point regarding displacment and strain gauge. I am a bit stuck with my accelerometer sensor data. I am attaching the acceleroemeter data file, sampling frequency: 200 Hz. The data still shows a lot of noise. Can you please help me out.

Sign in to comment.

Community Treasure Hunt

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

Start Hunting!