Can I Split Frequency Ranges in Harmonic Analysis for Separate Simulations?

Hey everyone,

I’m working on a harmonic analysis simulation and have a question about splitting the frequency range.

Let’s say the frequency range I’m analyzing is from 10 Hz to 100 Hz. Instead of running the harmonic simulation for the entire range in one go, is it feasible to split the range into smaller intervals, for example:

  • One simulation for 10-50 Hz, and
  • Another for 50-100 Hz?

I’m trying to understand if this approach would still provide accurate results or if there are any drawbacks, particularly at the boundaries where the ranges are split. Has anyone used this method in their analysis, and if so, were there any issues with accuracy or continuity?

From my understanding, harmonic analysis calculates the steady-state response at a specific frequency. For example, consider the system at 10 Hz, assuming no damping. We know the equation is:

M(x′′)+k(x)=Fsin⁡(ωt)M(x’') + k(x) = F \sin(\omega t)M(x′′)+k(x)=Fsin(ωt)

Since 10 Hz is the starting frequency, the initial displacement at all nodes would be zero, x(t)=0x(t) = 0x(t)=0. By solving this, we get the response for this frequency.

Now, when the solver moves to 11 Hz, will it assume the initial displacement is zero again, or will it consider the solution from the previous frequency?

I’d appreciate any insights or suggestions regarding the feasibility of splitting the frequency range and how solvers handle initial conditions between different frequencies.

@ChandraHarsha @rjainu

1 Like

Hi @sgundu, thanks for posting on the forum

To my evaluation, you shouldn’t have a difference in result as every frequency would correspond to a different excitation of the model. I’ll check with the team and get back to you later.

Best,
Igor

Hi @sgundu, I’ve performed a simple test here where I ran an analysis for a single frequency and compared it with one that included such a frequency on a frequency list. Since the results were the same I think it’s safe to assume they previous frequency bears no influence on the next one (they are run as separate analysis, as expected).

Thanks for the response @igaviano.

I just reviewed the test, and I’m wondering—do you think we would get comparable results for model-based harmonic analysis? My thought is that if the eigenfrequencies are consistent across both runs, the results should be similar. However, if we miss some eigenfrequencies in one of the runs, the outcomes might diverge. What’s your take on this?

Thanks

Hi @sgundu, apologies for the late reply here!

I’m not sure if I follow what you mean, could you clarify a bit further?

As the harmonic solver is linear, I guess the inserting or not of some frequencies should not affect the results for any given frequency.

Best,
Igor

I see… Thanks