Error of the Ansys-CFX Fourier Transformation, in Transient Blade Row model
 

Dear all,
I'm faced up with a problem that keeps troubling me these days. I've tried all I can do but it didn't help.
When I'm applying Fourier Transformation to calculate the unsteady flow field of a single stage shrouded axial turbine, the CFX solver called an error that
+--------------------------------------------------------------------+
| ERROR #999100267 has occurred in subroutine gKCsFCCsPS. |
| Message: |
| |
| Unable to compute interface connectivity between the following |
| domain interfaces: |
| |
| Phase Corrected domain interface : rotpass |
| Sampling domain interface : rotpassin |
| |
| for this serial/partitioning run. Likely causes are: |
| |
| 1) The mesh has been transformed using single precision |
| arithmetic at some point. |
| 2) The single precision solver/partitioner is being used instead |
| of the double precision solver/partitioner as recommended for |
| the Fourier Transformation model. |
| |
| If 2) is not the case please increase the expert parameter |
| "ps mapping check tolerance" to a value larger than the current |
| value: 1.000E-012. |
| |
+--------------------------------------------------------------------+

+--------------------------------------------------------------------+
| An error has occurred in cfx5solve: |
| |
| The ANSYS CFX partitioner exited with return code 1. |
+--------------------------------------------------------------------+

The second passage is generated by Turbo Rotation, and the connections are fine. In order to make sure, I exported two-passage mesh in the grid generation software, but the error still happened.
After reading the advice on other blogs, I'm sure that the CFX is opened directly in the system, for which the problem might not be caused by CFX(beta), or the beta component is referred to the .beta variable of CFX-post? And I don't know how to switch off the calculation of .beta variable (Density .beta, for example), even after I read the user manual.
The steady result which is used as initial files are calculated as double precision as well.
However, I enlarged the "ps mapping check tolerance" in the CFX-Pre to 1E+12, but the problem still happened.
Can anyone helped me with the problem? Really really thank you.