pipe_of_the_living_dead
This repository holds all materials for Team Hashmap from the 2019 Devon Hackathon. The project attempts to automatically identify and accurately count the number of well pipe segments from an image.
Business Context
It's considered a critical task to keep accurate count of tubulars on a well site and since pipe can't be counted in the hole, it's required to keep inventory out of the hole compared to the pipe on location. The field teams use this as a method of verifying depth compared to sensor readings so that they can ensure enough tubulars are on location to complete a job.
Today's Process
Today, 2 people do a manual count of the pipe segments and then compare results which takes a significant amount of time. Attempting to move away from the manual count has significant challenges though including:
Why Is This Difficult
-
Individually tagged pipe not feasible (RFID tag durability issues, high $, difficult to justify)
-
Pipe racks are not even or uniform and the pipe is layered
-
Different pipe locations in rack (hiding in shadows based on the layering and uneven stacking)
Challenge Specifics
Application / Solution Requirements
-
Runs on a mobile platform
-
Provides an accurate count (including pipe in shadows!)
-
Simple and easy to use
-
Captures different "racks" of tubulars including adding them together or keeping them separate
-
Keeps a history of "counts" with "rack" or description
Bonus Points For:
-
Add entry for average joint length by multiplying by count to get total length of tubulars
-
Dimensional calculator (approximate outside diameter of pipe)
-
Classify how many pieces have thread protectors installed