Sorry for the misunderstanding...it's not very easy sometimes to figure out the nature of our problems...never mind...
About your last post:
My Procedure:
1. Run Modal Analysis
---> It is not necessary for harmonic analysis (motion equation is directly solved...)
2. Set up an Harmonic Analysis - leave settings at default
---> ok
3. Apply a Structural Force to the attachment nodes in the X direction.
---> Why only in one direction? Solution must be the same when you apply Fx, Fy and Fz simultaneously.
4. Write Load Step
---> not necessary if Fx, Fy and Fz applied simultaneously in the same load step.
5. Repeat for the Y and Z directions
6. Load Step Opts >> Time/Freq >> Damping >> DMPRAT=0.02
--->OK
7. Load Step Opts >> Time/Freq >> Freq >> Ramped >> 3 substeps
--->So, You want to analyse 3 frequence responses? If you want to analyse a 1 single frequence response (to identify your problem precisely) you may set Load Step Opts >> Time/Freq >> Freq >> Stepped and set one step between 0 and the specific frequence from wich you need to compare other results.
8. Solve >> From LS Files
9. Create Loadcases for all substeps
10. For each frequency, read Loadcase FX, Square it, add Loadcase FY^2, add Loadcase FZ^2, Squareroot
---> Are your output Items summable (see 5.5.5.3 of ANSYS V9 online manual)? If yes, do you combine both Real and Imaginary part? Could you be more precise. Normaly, harmonic results contain one real part and one imaginary part (to be combined in order to view amplitude result. See next point).
11. Check Results
--->If you are checking results remember to set your phase angle to a value > to 360° to chek amplitude result(HRCPLX,load step, sub step,400) but you are probably aware about this.
If your are aware about all this points, we should probably search another origin problem. Example: When 2 modal frequencies are close enough, they can lead to a one single peak response (instead of 2 distincts)...to see...
Just one more thing about : "Sanosan, from what you say, I think that the stresses will never match, but the mode shapes surely should?"
When you analyse a single frequence response, your deformed shape don't match necesseraly the mode shape of the correponding modal analysis. It depends on the configuration of your loading. Does your load excite the researched mode shape (as stringmaker said it in his last post). Moreover, even if your load excites the strcuture at the mode frequencies requiered, other modes can participate to the response. Your load can excite multiple modes so that the result mode shape is a combination of multiple modes which are excited too (in general with lower participation). In this case mode shape don't match.
I wish my discussions will be helpfull for your analyse.
Good work !