I have never used Ansys but I wrestle with MAC every now and then.
MAC is nothing more than a comparison between modeshapes. Say, you measure a set of modes and store these as a vectors Vi, where i is 1, 2, 3 to N You then compute a set of modes Uj , where j is 1, 2, 3 to M
MAC is defined as
MACij = |ViVj*|2/(|Vi|2|Vj|2) ,
where * is complex conjugate and | | is absolute number.
If you are used to signal analysis, you may see that this definition is based on the same reasoning as for Coherence. http://en.wikipedia.org/wiki/Coherence_%28signal_processing%29
When modeshapes Ui and Vj are identical, MACij is unity and when the modeshapes are orthogonal, the MAC value is zero.
Comparing Test/Simulation, you can never assess all positions and therefore, your MAC values will never be perfectly unity or zero. As a rule of thumb you settle for MAC values of 0.8 or higher as identical modeshapes and 0.15 or lower for 'orthogonal' mode shapes.
RSTMAC seems to be Ansys talk implying RST (for its ReSulTs output, i.e. its binary output data) and MAC.
To get your RSTMAC to work, you must have two sets of modes. One computed and another measured and/or computed.
As you can see from the above definition. It is very easy to calculate the MAC, so you can just as easily do it in Excel, Matlab or similar.
If you are going to update a simulation model to test results, you may want to take a look at my ramblings here. http://qringtech.com/learnmore/why-simulate-measure-correlate-automate/
I have never used Ansys but I wrestle with MAC every now and then.
MAC is nothing more than a comparison between modeshapes. Say, you measure a set of modes and store these as a vectors Vi, where i is 1, 2, 3 to N You then compute a set of modes Uj , where j is 1, 2, 3 to M
MAC is defined as
MACij = |ViVj*|2/(|Vi|2|Vj|2) ,
where * is complex conjugate and | | is absolute number.
If you are used to signal analysis, you may see that this definition is based on the same reasoning as for Coherence. http://en.wikipedia.org/wiki/Coherence_%28signal_processing%29
When modeshapes Ui and Vj are identical, MACij is unity and when the modeshapes are orthogonal, the MAC value is zero.
Comparing Test/Simulation, you can never assess all positions and therefore, your MAC values will never be perfectly unity or zero. As a rule of thumb you settle for MAC values of 0.8 or higher as identical modeshapes and 0.15 or lower for 'orthogonal' mode shapes.
RSTMAC seems to be Ansys talk implying RST (for its ReSulTs output, i.e. its binary output data) and MAC.
To get your RSTMAC to work, you must have two sets of modes. One computed and another measured and/or computed.
As you can see from the above definition. It is very easy to calculate the MAC, so you can just as easily do it in Excel, Matlab or similar.
If you are going to update a simulation model to test results, you may want to take a look at my ramblings here. http://qringtech.com/learnmore/why-simulate-measure-correlate-automate/
Thank you so much Claes and Naveen. I have good idea about what MAC is. My concern is specific about using ANSYS to get the MAC values and play with it. Claes, your reference is helpful. Thank you.
If you really want to use ANSYS (for comparing modal basis from different .rst):
Documentation: help/ans_cmd/Hlp_C_RSTMAC.html
In a few steps:
1- prepare a directory architure such as in attached png: directoryArchi.png,
Here I have 3 .res:
- stsp1: plate model, pinned edges, all possible modes extracted (no matter it make sens or not),
- stsp2: plate model, clamped edges, all possible modes extracted (no matter it make sens or not),
- stsp3: plate model, pinned edges, truncation at 20 modes (hope it makes sense this time :) ).
2- use the command in post1 post-pro as shown in attached png: command.png ,
3- you obtain the kind of results as shown in attached .txt: macResults.txt ,
4- play with options depicted in documentation to obtain the analysis you wish.
In my .txt, you will see that the MAC between each pairs of stsp1 and stsp3 eigenmodes is more or less perfect (as expected). I will let you find your own conclusions for stsp1 and stsp2 (in the second pass, I sued maclim=0 to obtain a full matrix) ...