Q: Applying inertial loads in transient mode-superposition problem

[NCR9026]

I'm using ANSYS 9.0 to model the deformations of a flexible, rotating wing subject to inertial loads caused by the rotational motion. I have been using the full nonlinear TRANS solver successfully. I'd now like to run the same simulation using the mode-superposition solver and compare the results, but I'm having difficulties transitioning between the set-up processes between the two.

In my non-linear model, I have been applying inertial loads via the OMEGA, DOMEGA, and ACEL functions at each time step:

TIME,...        ! set time
OMEGA,...       ! set angular velocities
DOMEGA,...      ! set angular accelerations
ACEL,...        ! set linear accelerations
LSWRITE,...     ! write load step

My input velocities and accelerations are known analytically as functions of time. Each load step has non-zero velocities and accelerations in all three axes. I define as many load steps as necessary, then run LSSOLVE.

In contrast, my understanding is that if I want to apply inertial loads for the mode superposition method, I have to apply a load during the initial mode solution. Then the LVSCALE command is used scale only that defined load as the transient solution progresses, correct?

Since I have 9 inertial loads (3 rotational velocities + 3 rotational accelerations + 3 linear accelerations), it seems like I'd have to run the mode superposition solution 9 times---once for each inertial load vector component---using LVSCALE to independently control each load, then sum the results of all 9 runs to get the total response.

Is the approach I've outlined correct? Is there any easier way to set up this problem?

Thanks.

 

[NCR9026]

Since I have 9 inertial loads (3 rotational velocities + 3 rotational accelerations + 3 linear accelerations), it seems like I'd have to run the mode superposition solution 9 times---once for each inertial load vector component---using LVSCALE to independently control each load, then sum the results of all 9 runs to get the total response.

I've given some more thought to my problem and realized the above approach will not work, because it will not properly account for motion-dependent stiffnesses (e.g., centrifugal stiffness).

The only other approach I can think of is to continuously stop and re-initialize the analysis at each time step, which seems cumbersome to say the least.

Can anyone else give any other suggestions?