The fact a natural frequency is overestimated or underestimated in MANATEE can come from different reasons:

- wrong value of the winding / lamination weights -> check the outputs of MANATEE and the input geometry (or use a post-process like plot_weights or plot_machine)
- difference in radial stack stiffness due to the presence of rods or welded plates on the lamination, or due to the frame (different shrink fit, presence of ribs)
- non circular (typically square) lamination
- 3D effects not accounted in the 2D analytical model if you don’t use MANATEE FEM tool

In case you have too high differences, the possibilities are the following:

- enforce the experimental natural frequencies using the variables
**Input.Simu.ExpFreq, ExpModes, ExpDamp**input values with**Input.Simu.is_force_natfreq = 1** - use the tuning factor
**Knat_freq_user**to tune the 2D analytical model - these tuning factors can account for the difference between the model and the real structure - use the 3D FEM structural model (longer calculation time but more accurate results) of MANATEE

It is anyway recommended to carry an experimental modal analysis of the electrical machine, as the manufacturing process can affect the natural frequencies and the damping, and use these experimental values to tune MANATEE model and have more realistic results. You can contact EOMYS if you need help to run this type of experimental characterization.

Note that this methodology is not necessary in a relative design analysis at variable speed (ex: influence of the slot number on the maximum sound power level): as resonances occur usually far from the minimum and maximum speeds, a +/-10% change of the natural frequencies will not change the largest noise and vibration levels.