# Band pass filter / RMS filtering

Phil Gooch

Posted on 21.01.2017 22:46

I am looking for some assistance interpreting the requirements of a document in relation to DeweSoft filtering.

Attached you will find some data that we recorded on a rail vehicle using a tri-axial accelerometer. We now need to evaluate the data according to the following:

"Accelerations obtained for lateral instability evaluation should be sampled at no less than 50Hz before being filtered band pass filter set to ±2Hz to the dominant frequency, then a sliding root mean square value of acceleration is calculated for a window length of 100m and a step length of 10m."

I have the following questions:

1. I know how to apply a band pass filter but how do I set it to +/-2Hz to the dominant frequency?
2. Can you tell me how I should calculate the second half of the requirement? (a sliding root mean square value of acceleration is calculated for a window length of 100m and a step length of 10m)

The criteria for this test is "A vehicle during lateral instability evaluation shall not experience a root mean square value of lateral underframe accelerations that exceeds 0.25g over any 100m window. "

Attached files:
Luka Jerman
Software Engineer II
Posted on 25.01.2017 09:57

Hello Phil,

1. I know how to apply a band pass filter but how do I set it to +/-2Hz to the dominant frequency?

We assume that you are having problems on how to determine the dominant frequency. There are two options inside DEWESoft math you can calculate the Fourier Transform of the signal or determine the exact frequency with an Exact Frequency mathematics. This two methods directly couldn't produce the desired output due to the characteristics of your signal.

You've mentioned that acceleration has to be sampled with no less than 50Hz. Therefore you probably need to observe frequencies that are lower than 20 Hz. The frequency domain diagram that was calculated by the FFT doesn't reveal any significant peaks at lower frequencies, because of higher frequencies that are a consequence of vibrations.

We observed that there are repeating shocks in your Acceleration Y signal. With DEWESoft Basic statistics, we calculated the peak to peak amplitude of the signal for every 0.01s block and performed an FFT. A peak can be seen at around 16 Hz (it varies in frequency and amplitude due to the changes in velocity).

However we would suggest that you find out if there is a standardized approach to finding out the dominant frequency. Usually filtering procedures are exactly specified (filter type, cut-off frequencies...).

2. Can you tell me how I should calculate the second half of the requirement? (a sliding root mean square value of acceleration is calculated for a window length of 100m and a step length of 10m)

This is a bit difficult since the vehicle isn't driving at constant speed. We made the following steps to calculate this:

1. To get the distance we integrated the velocity which was previously transformed to m/s,

2. then we made a formula that increases for 1 for every 10 meters traveled but when it reaches the number 9 it goes back to 0,

3. we then made 10 triggered RMS calculations that trigger depending on the value in a formula that goes from 0 to 9 (2.). Therefore one is triggered when the formula goes from 0 to 1, the next one when it goes from 1 to 2....

4. the previous triggered RMS channels were combined into one channel that has a value of first RMS if formula (2.) has a value of 1, it has a value of second rms if the formula (2.) has a value of 2....

This method involves some additional channels but it seems to be most straightforward.We've added the mathematics in your data file you can review them on screens DEWE_AccY and DEWE_RMS.

File is available on the following link: