Cancel
Showing results for
Did you mean:

# Modal Tips: Roving Hammer versus Roving Accelerometer

Siemens Phenom

Modal Tips: Roving Hammer versus Roving Accelerometer

“To rove or not to rove, that is the question” - Anonymous

When performing a multiple point modal test with a force impact hammer and accelerometer, there is a choice to move or “rove” either the hammer or the accelerometer between measurement locations.

Roving the impact hammer has some advantages over roving the accelerometer.  When moving the accelerometer, one has to unmount and remount the accelerometer at the different measurement points.  Moving the accelerometer also changes the mass distribution on the structure, which could alter the natural frequencies.

Moving the hammer is not without considerations however.  Because the hammer is a single input and a triaxial accelerometer measures output is three directions, one must take care to ensure the resulting mode shapes are complete.

Question:

I am doing a modal impact test on a structure with an impact hammer and a single triaxial accelerometer.  There will be nine measurement locations as shown in Figure 1.

Figure 1: Modal test setup (with impact hammer and triaxial accelerometer) and structure (with nine measurement locations)

I want good mode shapes.  Does roving the accelerometer versus roving the hammer make a difference in the modal results?

Yes! It can make a difference in the mode shape results if you are not careful.  It is possible to perform the test and have incomplete mode shape information.

The accelerometer measures in 3 directions and the hammer measures in only one direction.  If you move one versus the other, this difference needs to be taken into account.

Background

When performing a modal analysis test, consider the possible input references and possible output responses associated with the different measurement locations.  A table of the possible inputs and outputs for a nine location test is shown in Figure 2.

Figure 2: For a nine measurement location modal test, there are 27 possible inputs and 27 possible outputs that can be measured.

When measuring the Frequency Response Functions (FRFs) between the nine measurement locations:

• There are 27 possible input locations for applying the input force. At any given location (1, 2, 3, 4, etc.) the force (F) can be applied in three possible directions: x, y, and z. Possible inputs are therefore F1x, F1y, F1z, F2x, F2y, F2z, etc.

• There are 27 possible output locations for measuring the acceleration response. At any given location (1, 2, 3, 4, etc) the acceleration (A) can be applied in three possible directions: x, y, and z. Possible outputs are therefore A1x, A1y, A1z, A2x, A2y, A2z, etc.

To have a proper mode shape, where the phasing of all the measurement points are aligned properly, the measurements must all have a common reference.

A common reference means that either a complete row or complete column of the table shown in Figure 3 must be measured.

Figure 3: A complete row or complete column of the measurement table must be measured to ensure consistent phasing for the mode shape.

A set of Frequency Response Function measurements corresponding to any single or row or column are required.

Roving Accelerometer

Consider the case where the accelerometer is moved between the nine different locations while the hammer is applied to the same location (F1z) as shown in Figure 4.

The accelerometer, which measures in three directions, is first located at measurement location 1.  The A1x, A1y, A1z outputs are measured at one time with respect to F1z.

Figure 4: Roving accelerometer outputs with fixed hammer inputs

After measuring the three FRFs for location 1, the accelerometer is then moved to location 2.  The three outputs of location 2 (A2x, A2y, A2z) are measured with respect to the hammer input at location 1 in the z direction.  This continues for locations 3 thru 9.

After all nine locations are measured, one complete row of the table is filled in as shown in Figure 5.

Figure 5: Roving accelerometer modal test has one complete row: all outputs measured with respect to one input

The complete row with common reference assures that the phasing of all the measurements are consistent, so that a proper mode shape can be calculated.

Roving Hammer

Instead of roving the accelerometer, the hammer could be roved while the accelerometer stays fixed.

In this case, the accelerometer measures output A1x, A1y, and A1z.  The hammer impacts at F1z, which happens to be perpendicular to the surface of the test object.  The hammer is then moved to F2z, F3z, etc.  This happens until impacts are done at all nine locations as shown in Figure 6.

Figure 6: Roving impact hammer inputs with fixed accelerometer outputs

The result of this test is NOT a complete row or column in the table after the test is complete as shown in Figure 7.

Figure 7: Roving hammer modal test DOES NOT have a complete row or column

There is NOT a complete row or column with the roving impact hammer test.  The impact hammer was only used in one direction at each different measurement location, instead of being used in all three directions.  This is an easy oversight when using a modal impact hammer.

When trying to use an impact hammer on a flat surface, it is very difficult to impact in other than the perpendicular direction to the surface.  This is why it is often easy to overlook the need to impact in all three directions when performing an experimental modal analysis.

If the impact hammer was used to input in all three directions (x, y and z) at each location, then three complete columns would have been created.  This would result in complete mode shapes as shown in Figure 8.

Figure 8: Complete mode shapes

There were 27 possible inputs and 27 possible outputs (9 locations, three directions each) in this modal test.  This means to get a complete row or column, 27 FRF measurement functions must be done at a minimum:

• With the roving accelerometer case, only nine actual measurements were required. Each time the accelerometer was moved, three measurements of the associated row in the table were completed.
• In the roving impact hammer case, twenty seven separate measurements must be done. For a given modal impact hammer input, only one measurement is added to a single column in the table.

If the modal impact hammer is applied in all three directions at each measurement location, a proper modal test would be done, with three complete columns of the table.

Conclusion

While roving a hammer versus roving an accelerometer may outwardly appear to be the same for a modal test, the results can yield an incomplete mode shape.  It is important to take care to make sure a complete set of outputs with respect to consistent input are measured.

Questions? Email peter.schaldenbrand@siemens.com or contact Siemens PLM GTAC support.

Be sure to check on the free On-demand Webinar Fundamentals of Modal Analysis.

Modal Data Acquisition

Modal Analysis and Operational Deflection Shapes

Other Articles:

• ,
• ### Structural Dynamics Testing

Dreamer

Hello,

Lets suppose we are testing a plate. In a roving hammer test impacting in Z direction and using only one accelerometer we can get the out-of-plane modal shape. Does it means that using a triaxial accelerometer in the same roving hammer test the in-plane modal shapes (X and Y) cannot be extracted?

Actually, I tested a plate-like structure and when I compared FEM in-plane modal shape (X and Y) displacements results with those I tried to extract from a roving hammer tests did not match.

Thanks for any help,

Siemens Phenom

Jesus,

Yes this is true.  The only way to get all 3 directions from a roving hammer test is to impact in X,Y,Z directions at each point of the structure.  This would provide a complete column in the FRF matrix matrix for a single axis accelerometer that does not move and three columns for a triax accelerometer that does not move.

Since it's difficult (or near impossible) to impact in 3 directions in the middle of the plate we normally suggest using roving accelerometers when 3 directions are required.  The other option would be to glue blocks at each node, impact the blocks in 3 directions and then use Modal Modification to remove the mass of the blocks at each node.

kevin

Contributors
Latest articles