Modelization of Yield with volumetric information (geographic coordinates and depth of salt mine)

Question

Hello,

we have customers extracting salt from mines using injectors and extractors. They have coordinates (latitude / longitude) for both injectors and extractors, depth of pipes, depth of the mine, thickness of the salt vein, distance to the edge of the vein, .... and they wants to modelize Yield accross time depending on extractor coordinates and aforementionned informations which seems to be a spatial (volumetric as depth is critical) modelization. Is that possible ? any example in the Gas&Oil field ?
Thanks for the help.
Emmanuel

Peter_Polito · Accepted Answer

@Emmanuel_Romeu , what you are describing is quite similar to problems we're dealing with in the oil and gas industry. Horizontal drilling is becoming the norm so the lat/long of the well head becomes less important than the lat/long (and depth) of the screened portion of the well. While different methods are used to extract the media, there seems to also be many parallels between salt extraction and oil/gas well completion parameters. The most popular models in O&G are all ML-based with preference going to neural networks and bootstrap forest. Neural networks are "hip" right now but I have had marginally higher success with bootstrap forest. There is growing interest in XGBoost but its kind of the new kid on the block and not widely known but might be useful in this instance. I have a growing number of O&G datasets if that would help.