Author: Piyush Modi / Source: Forbes
The oil & gas industry still faces serious challenges, including the costs of maintaining its aging infrastructure. On average, 42% of offshore equipment is more than 15 years old and has 13% downtime – the vast majority of which is unplanned. It is estimated that refiners in the US alone lose around $6.
6B a year due to unplanned downtimes.For more than a decade, the rallying cry has been to turn to massive data collection and analytics. The numbers are substantial. Pipeline inspection generates 1.5TB for every 600km inspected, ultrasound around 1.2TB for every 8 hours of scanning, process data collected is around 6GB per plant per day and seismic surveys generate around 10TB each.
Unfortunately, this data has not yet turned out to be “the new oil” many predicted. 60% of operators still cite delivering outcomes from data as a major problem. The industry struggles to unlock end-to-end insights from the data it has been collecting. More than 95% is never used, and hand-made analytics are too slow and don’t scale well.
The reliance on traditional analytics techniques, and a legacy CPU-compute infrastructure that lacks the needed processing power to analyze the volume and variety of data fast enough has also hampered progress. Oil and gas data needs algorithms and compute that can scale to distill these oceans of data, deliver insights and maintain efficiency.
The larger ecosystem of vendors has started to respond in the oil and gas sector. For example, companies such as MapD are using GPUs to provide in-memory database and real-time interactive visualization solutions benefitting from GPU-accelerated data aggregation to visualize millions of rows of data in real time, spanning drilling, production, and supply chains operations for better data utilization and timely decision support.
Deep learning, GPUs and the concept of “Digital Twins” offer enormous potential benefits for predictive maintenance in oil and gas. For example, early and accurate detection of faults, to predict remaining useful life of an asset given an operational context or to even prescribe guidance on work scope for the field service team with recommendations of parts and personnel skill desired to service them.
By definition, a “Digital Twin” is a continuously learning system of digital copies of assets, systems and processes that can be queried automatically, or even by voice, for specific outcomes. A digital twin can predict asset behavior and capacity to deliver on specific outcomes within given parameters and…
The post How AI Is Providing Digital Twins For Predictive Maintenance In Oil And Gas appeared first on FeedBox.