In general when you define seismic case you usually define it assuming that the excitation is in the positive direction of X or Y but of course in reality it can be in the opposite directions too.
True, but already considered on seismic combinations (+-X+-0,3Y and +-0,3X+-1Y)
This is why they are duplicated and half of them gets -1 factors.
Not true to axial forces on seismic combinations.
Please mind that for vibrations the same point on the structure can move in the opposite directions causing tension or compression. Imagine a simple portal frame with horizontal force applied from left and then from right (structure movement is right - left - right - left ...).
Right artur, but seismic combinations also have Death and Live Loads, both vertical loads.
You can only think in the way you have in that example to seismic action acting alone...where multiplying axial by -1 is correct. You cannot multiply axial force from vertical loads.
So i'm not seeing the way that this procedure on robot can be correct.
I'm afraid this wasn't well communicated from my side. I'll try again in the different way. Imagine a frame as on the picture I attached previously. You define earthquake in the direction X+. Let's assume that only one mode is used and this is the one that causes the structure to move from right to left and back again. This movement (swinging) is repeated number of times (vibrations). The seismic analysis will give you result for let me say max. amplitude of the vibration (in the time history analysis you have many time steps and different level of forces in the elements corresponding to the deformation of the structure at each of these steps) and taht is e.g. when structure moves right just before starting to moving back left. In this point you have compression in the right column and tension in the left one. So the vibrations continue and the structure moves left. Now you have the compression in the left column but tension in the right one. For the designed (e.g. right column) you need to multiply the results obtained for DirX+ case by -1 to include the fact that the frame moves in both directions.
I hope I did better this time.