This article explains:
After optimizing, analyzing and tolerancing a sequential optical system, it is often required to convert it to a non-sequential system, typically when needing to look at detailed, opto-mechanical stray light problems.
There is a convenient tool in sequential ZEMAX, under Tool>Miscellaneous>Convert to NSC Group, that automatically converts a range of surfaces to the equivalent non-sequential objects. This tool converts most commonly used surface types, surface apertures and coordinate breaks into non-sequential component group. However, some sequential surfaces do not have a non-sequential equivalent and cannot be converted. Also, there are configurations that cannot be converted or not yet supported.Carefully check the converted lenses when using this tool.
Once the lens has been successfully converted, it is easy to add objects modelled in CAD packages to represent mounts, baffles, iris apertures etc and look in detail at the interatction between the optical and mechanical components of the system.
The conversion tool capabilities are constantly being updated to support more surface types. For more detailed information about the currently supported surface types, please refer to chapter 8, section Miscellaneous>Convert to NSC Group, in the most current version of ZEMAX manual.
We will go though an example on how to do this conversion . We will use the Cooke Triplet sample file located in {ZEMAX root}/Samples/Sequentia/lObjectives/Cooke 40 degree field.zmx. 
The goal of this exercise is to convert from surface #1 to #6 to equivalent non-sequential components and place a non-sequential detector object at the location of the current image plane (surface #7). We will also place a non-sequential source that represents the beam on-axis in object space. This will let us confirm that the converted system works correctly.
After the conversion, ZEMAX will create a Non-Sequential Surface in the Lens Data Editor containing all the non-sequential components, thus making the system "mixed mode" (non sequential with ports). In a mixed mode system, rays are traced sequentially outside the non-sequential group (outside the Non-Sequential Surface type) and non-sequentially inside.
Now, the concept of a "Stop surface" only applies to sequential ray tracing. This is because in sequential ray-tracing rays are aimed initially at the entrance pupil, which is the image of the stop, or at the stop itself. Therefore, only a sequential surface can be set as the system Stop. The Stop surface must precede the non-sequential segment of the design.
In the Cooke triplet example, the Stop surface is embedded in the system. Consequently, we need to move the current stop location to a dummy inserted before the first lens we want to convert.
Also, the all semi-diameter should be fixed (“U” next to the semi-diameter) before converting to a non-sequential design. The Semi-Diameter values in this file are by default fixed.
Insert a dummy surface prior to current surface #1.
Set the dummy surface as the stop in the Surface Property window. To open the surface property window, double click on the "type" column of the dummy surface.
After clicking “OK”, the Lens Data Editor will display“STO” next to the dummy surface indicating that the surface is a stop.
Select the Convert to NSC Group tool in the Tools> Miscellaneous menu. 
Specify surfaces 2 to 7

We now have a mixed mode system, or non-sequential with ports as referred in chapter 12 of the manual, indicated by the non-sequential surface in the Lens Data Editor. For details about the difference between non-sequential with (mixed mode) or without ports (pure non-sequential mode), please refer to the first part of chapter 12 in the manual.
For this exercise, we will convert the file to a purely non-sequential one rather than leaving as a mixed mode file. Click “non-sequential mode” under File in the main menu.

Click OK in the following dialog box

Once converted to pure non-sequential mode, the Non-sequential Component Editor (there is no Lens Data Editor in pure non-sequential mode) will contain non-sequential components corresponding to the lenses in sequential Cooke Triplet file.
Open the 3D layout under Analysis>Layout>NSC 3D layout.
The lens appears to have converted correctly!
The sequential system had an entrance pupil diameter of 10 lens units. To create the same on-axis input beam, we can place a collimated, circular non-sequential source object of the same size to the left of the first lens.
Insert a new line anywhere is the non-sequential component editor.

Open the Object Property window by double clicking on the Object Type column in the non-sequential component editor and set the type as “Source Ellipse” and click OK

In the editor, set the following parameters for the Source Ellipse and leave all other parameters same as default.
Z position = -10 (since it is collimated, it does not matter where it is located as long as it is to the left of first lens)
# Layout Rays = 10
# Analysis Rays = 100000
X Half Width = 5
Y half Width = 5

Update the 3D Layout; it shows the source with 10 analysis rays.
Rays are tracing correctly!
Now we need to place a detector object at the same location at the sequential image surface. The sequential image surface was at a distance of 60.177 units to the right of the first surface in the Lens Data Editor. Since in the non-sequential file the first lens (object #1) is located at the global origin (i.e. Z position parameter is zero), we need to place the detector at Z position of +60.177.
Repeat the previous step and set the object type as “Detector Rect”.

Here are the parameters for the “Detector Rect“ object
Z position = 60.177
X half Width = 0.01
Y half Width = 0.01
# X Pixels = 100
# Y Pixels = 100
The updated layout will change as follow.
Open the Detector Viewer by clicking Analysis>Detector>Detector Viewer. For now, the detector viewer will show a blank window. We need to trace analysis rays to the detector to see the optical power landing on the detector.
Open the Detector Control window by clicking Analysis>Detector>Ray Trace / Detector Control.

Hit “Clear detector” first, to clear the content of the current Detector Viewer, and then press Trace followed by Exit. ZEMAX will trace 100000 rays as specified in the "# Analysis Rays" parameter of the Source Ellipse in the editor.
The Detector Viewer now shows the irradiance distribution.
If you do not see the same output, make sure that the settings for the Detector Viewer are as follows.
The distribution shown in the Detector Viewer corresponds to the sequential on-axis geometrical spot diagram. You can open the Spot Diagram in sequential ZEMAX to compare the result.
So far, we have compared the geometrical ray trace results between sequential and non-sequential file. We can also make the comparison between diffraction calculations, specifically Huygen’s point spread function. Set the following parameters for the source and detector in the editor.
Source
# Analysis Rays: 3000 (reduced to speed up the detector trace)
Detector
Data Type: 1
PSF Wave#: 2
Also, set the Detector Viewer to coherent irradiance.
After tracing analysis rays to the detector it will display the Huygen's diffraction point spread function. 
Here is what the equivalent sequential analysis looks like:
Both sequential and non-sequential analysis shows practically the same results. The small difference in the intensity of the second ring is due to the different number of rays traced.
Now that we are confident with the conversion process, we can carry on and do further modeling with the non-sequential file. You are encouraged to learn about non-sequential ray tracing in ZEMAX by reading related knowledge base articles, studying the sample files and referring to the manual (chapter 12).
This article has demonstrated how to convert sequential surfaces to non-sequential components using the conversion tool. In summary:
Further Reading
1. ZEMAX User Manual, Chapter 12