While the "best" method is probably unanswerable since it would be based on very specific requirements and subject to change as soon as a better method were devised, here are some feasible methods and some issues with using the print head:
I believe three methods may prove effective in achieving auto-ejecting parts from consumer or entry-level professional 3d printers. I am not aware of these strategies having been practically demonstrated, but I have devised (not built) these methods as options for my own printers. I'll update this if/when I (or others) demonstrate any of these methods.
Scrolling Conveyor-type Bed:
- In this concept, parts become dislodged from the print surface as it is deformed around a roller in the process of scrolling to the next position. Scrolling bed designs must make allowances to prevent parts from lifting up the bed material which becomes an issue especially with warp-prone materials. Note: This is the basis of the Automated Build Platform (ABP) originally designed (as far as I can tell) by Charles Pax and later covered in several patents by Makerbot Industries.
Deforming Bed:
- In this concept, the bed is mechanically deformed when the part removal temperature has been reached. This deformation dislodges the part which can then be easily swept off of the bed by an arm or similar mechanism.
Eject and Replace Bed:
- This method ejects the entire bed surface along with the finished parts and then receives a fresh print surface for the next print. This method would likely still require intervention to remove parts from used print surfaces and then return them to the clean stack.
Issues using print head to eject parts: While this method has been tried, and demonstrated (see below), it is not without challenges/drawbacks.
3d printers are not designed to apply significant force behind print head movements. While a printer designed specifically for this purpose could conceivably be built, it would require a substantial redesign and potentially require more expensive components.
Applying enough force to remove a part with the print head is extremely likely to cause the stepper motors to loose steps and result in loss of position accuracy. (however, position could easily be regained by zeroing via limit switches between prints.) A printer is not likely to see belt ratcheting (skipping) before skipping steps, but this would be the next likely result. This issue would be less likely for easily removable parts.
Real Examples: The above methods are theoretically sound, but have not been demonstrated (as far as I know). Until they are, the following real examples have been demonstrated and may be suitable for specific requirements:
Sweep parts off with frame and moving bed like this.
Ramming part with robust print head like this.
Ramming easy to remove part with print head like this.
Interesting question. I hope this helps!