We planned to upgrade the extruder from a nema43 stepper motor with a 1:20 gear to a servo motor with a 1:120 gear. The servo ran fine on the desktop, but when I installed it in the printer, it only moved sometimes and when it did move, it was very erratically.
We used the old stepper motor, risking that it would stall if there were too much material in the extruder, or the material was too dry. This happend twize during the print, but the damage was small enough that we could fix it with masonry tools.
The first version of the printer had the print-head itself moving up and down. This has always worked well, but with a dynamic moment of only 1.25m. The upgrade would be to build four large colums and have the whole printer moving on the colums. This failed because one of the servo motors had an internal encoder error, and we were not able to get a new one during the weekend.
We fixed the z-axis carriages with m8 bolts and used the old z-stage.
The plan was to use a m-tec duo-mix pump to mix the concrete and pump it directly into the material buffer in the print-head. When we started printing, the pump failed and we had to use buckets to fill the print-head. 3 fillings pr. layer, so a total of 150 buckets was used to print the one cubic meter of material used for the foundation.
The BOD was sliced with Simplify3d. I made a post-processor for it that would add the gcode to rotate the printhead to always be tangential to the print direction.
It sliced all our tests fine, but the foundation failed. The slicer would freeze at "Calculating infill" although we had no infill. The solution was to scale the model to 1/10 the size and have the post-processor scale the gcode back to real size.