Operational Modal Analysis OpenFAST

[Alessandro5I]

Hi,

I am working on implementing an OMA (Operational Modal Analysis) starting from an OpenFAST simulation. I have almost successfully done this with ElastoDyn, but I believe it would be more beneficial to implement it using BeamDyn instead.

I would like to know how to simulate accelerometers in BeamDyn. Specifically, is there a way to set the output nodes at precise locations along the blade? I attempted to modify the keypoints, thinking that the outputs would be at these keypoints, but it didn’t work as expected.

What do you suggest as the best approach to achieve this?

Thanks in advance,
Alessandro

[jjonkman]

Dear @Alessandro5I,

BeamDyn outputs along the blade like acceleration are output at nodes. If you are using Gaussian quadrature (quadrature = 1), these nodes correspond to the FE nodes (whose location is determined only by order_elem); if you are using Trapezoidal quadrature (quadrature = 2), these nodes correspond to quadrature points determined by the stations in the BeamDyn blade file, plus possible refinement (refine > 1). If you have specific locations where you want to output the response, I would use Trapezoidal quadrature and place stations at the exact locations along the beam that you want.

FYI: Key points in BeamDyn define the reference axis/curve, but are not directly used as analysis nodes.

Best regards,

[jjonkman]

Hi @Alessandro5I,

The nodes used in Trapezoidal quadrature in BeamDyn need not be evenly spaced. Rather, they are defined as eta between 0 and 1 along the curved reference axis in the BeamDyn blade input file, along with the mass and stiffness matrices at each node.

BeamDyn does not track points offset from the reference axis along the chordline, but these could be tracked (assuming rigid body motion for the offsets relative to the reference axis) through knowledge of the translation and rotation of the reference axis.

Best regards,

[Alessandro5I]

For now, I will proceed as you suggested. After selecting the blade station location, I will interpolate the mass (m), damping (c), and stiffness (k) matrices? To focus the output solely on the blade stations, what specific inputs do I need to set? Currently, I have 26 blade stations but 51 nodes in the output.

Additionally, is it possible to define the reference axis with different y-values to cover more points along the chord? For example, could I define two different y-values for the same z-coordinate to cover different points along the chord?

It would be interesting to have the output at different chord locations in the future, especially for structural health monitoring simulations.

Thanks,
Alessandro

[jjonkman]

Dear @Alessandro5I,

Yes, you can interpolate mass and stiffness matrices if you are adding new stations.

If you are specifying station_total = 26, but have 51 Trapezoidal quadrature nodes, this implies that you've set refine = 2, which splits each segment between adjacent stations into 2 (26 + 25 = 51). BeamDyn will linearly interpolate the mass and stiffness matrices between stations to find the matrices at the node in the middle of each segment.

There can only be one reference axis in BeamDyn and you can't specify more than one node at a given radial station. I agree that would be a nice feature to have in the future. Until then, tracking other points along the chordline should follow what I recommend in my prior post--through knowledge of the translation and rotation of the reference axis assuming rigid body motion for the offsets relative to the reference axis.

Best regards,

[Alessandro5I]

Hi @Alessandro5I,

The nodes used in Trapezoidal quadrature in BeamDyn need not be evenly spaced. Rather, they are defined as eta between 0 and 1 along the curved reference axis in the BeamDyn blade input file, along with the mass and stiffness matrices at each node.

BeamDyn does not track points offset from the reference axis along the chordline, but these could be tracked (assuming rigid body motion for the offsets relative to the reference axis) through knowledge of the translation and rotation of the reference axis.

Best regards,

Dear @jjonkman,

I plan to begin with the rigid body approach for estimating the accelerations. However, do you think it would be possible to later use the 6x6 matrices in BeamDyn to obtain a more precise solution for the accelerations at points offset from the reference line (keypoints line)?

Thank you for your insights!

Best regards,
Alessandro

[jjonkman]

Dear @Alessandro5I,

The 6x6 mass and stiffness matrices do not contain information that would allow you to calculate acceleration of points offset from the reference line. The underlying information (such as cross-sectional geometry and warping functions) from a sectional analysis used to derive the mass and stiffness matrices would contain such information, but this information is not known by BeamDyn.

Best regards,

[Alessandro5I]

Dear @jjonkman,

Thank you for your support.

I was wondering if there is a way to implement a strain-based Operational Modal Analysis (OMA) in OpenFAST using BeamDyn. In your opinion, would this be feasible? If so, how would you suggest approaching its implementation?

Thank you,
Alessandro

[jjonkman]

Dear @Alessandro5I,

Other than the fact that BeamDyn does not directly output strain, I would say "yes". You could swap strain for deflection in the OMA, which BeamDyn can output, or compute strain in a post-processing step via sectional analysis based on the section loads output from BeamDyn.

Best regards,