Artificial Intelligence: AI is Making Oil Rigs Safer
The energy industry is deploying AI-based, human-independent safety systems on oil rigs.
Oil rigs are one of the most dangerous places to work for human beings with a constant risk of injury or harm. Huge mechanical systems mesh together to keep these huge rigs afloat above the ocean surface and continue drilling deep beneath the ocean floor. The risks of fire, explosion, accidents from hazardous machinery, and falls are always round the corner on oil rigs. Artificial intelligence (AI) can make these danger-prone work areas much safer. (analyticsindiamag.com)
How Dutch company Rolloos is using AI
Rolloos supplies CCTV, communication networks, crane safety, data analytics, and other solutions for safety and security in the oil and gas industry.
Over four years ago, a worker was killed on a rig belonging to a client of Rolloos. A piece of equipment failed and a heavy pipe broke loose. It hit a worker standing nearby, known to be a dangerous place. The accident occurred due to a combination of mechanical failure and human error.
Bram Masselink, managing director of Rolloos, said he was shocked by the incident. It led to the creation of the “Red Zone System,” an AI-powered safety net for workers on rigs.
The system combines a network of cameras and sensors along with AI-powered computer vision algorithms run on Nvidia (NASDAQ: NVDA) GPUs. It immediately senses and locates rig workers in dangerous locations and issues an alert warning before the occurrence of an accident.
Red Zone System
Rolloos and the client worked together to develop the Red Zone System. They commenced by mounting cameras on the rig and capturing various types of operations performed in all kinds of weather and conditions over two months.
They then got human annotators to mark persons and objects in the video data. The team used this annotated information to build an algorithm that could sense people in various locations of the complex rig environment. It could raise an alarm if it sensed any danger to personnel.
After training the model on the label data model has been able to pinpoint locations within 18 inches. It can run a complete sequence in just 250 ms. The AI team uses false alarms to train and refine the model further.
The system runs on local hardware and not on the cloud. That’s because of the unavailability of Internet connectivity on rigs.
Rolloos trained its deep learning models on NVIDIA data center GPUs and deployed them for inference on an NVIDIA T4 Tensor Core GPU
Related Story: Is the Oil and Gas Industry Slow on the AI Uptake?
Latest Alternative Investment News
The inaugural suite of five ETFs from Guardian Capital comprises two Directed Outcomes solutions that combine high conviction ideas with derivative overlays to seek high income and risk mitigation and…
Gong, the startup which helps businesses glean “revenue intelligence” from customer interactions, achieved unicorn status when it raised $200 million at a valuation of $2.2 billion. Coatue led the Series…
Apollo Fintech announced Wednesday the completion of its National Payment Platform (NPP), the first of its kind, blockchain-based, cashless national currency platform. It enables a central bank to issue a…
Revolut, the UK-based neobank led by Nikolay Storonsky, on Tuesday called an end to its beta mode operations in Australia. Country CEO Matt Baxby said the fintech was now readying…