Re: Drones - not in the sense I think you are imagining, but you can pull data from some of the mining equipment and I have seen 3D scanners mounted in situ at test sites to capture specifics. (it didnt work well - got remodelled later).
The general workflow I have seen would be familiar to anyone in game or animation pipelines. A team models from the mine layout map in traditional 3D tools (e.g. Blender, 3DS Max) this may be combined with telemetry information from cutting machines (distances, seam heights, cutting rates etc) and then add additional visualization and environmental affects in game engine such as Unity or Unreal for example adding particle systems for dust etc. There are also industry specific specialist simulation and training engines.
If you are working with an equipment manufacturer, equipment models are usually converted from CAD data, then UV mapped, textured and keyframed either in traditional 3D tools or use programmed procedural animation in engine. Though you can purchase equipment model assets - features have to match depending on training fidelity. Equipment control is usually matched to actual physical controllers - so you read the input from a real world controller and trigger the corresponding virtual model animation in the virtual environment. The mining equipment manufacturers have training teams who create the assets and could configure layouts for specific client scenarios. As KineticLensmen pointed out - its the last stage of customization for specifics that usually takes longest since its crucial for the spatial awareness of teams being trained.