(99a) Machine Vision Safeguard Management

With the drive for increasing automation, unstaffed or lone-staffed facilities, and improved process safety performance, new technologies are needed to assure that safe operations are being maintained. While typical practice has been to include more increasingly complex sensors, single purpose sensors can face limitations, especially at the human/machine interface. The transitory nature of high-risk work and the difficulty with adding sensors to workers can leave this critical work as a blind spot to our engineered systems. As such, we often rely on administrative or procedural controls to ensure that safeguards are in place and work remains safe.

With the advancements in AI capabilities provided by machine learning, there are an increasing number of objects and actions that can be recognized and monitored by a simple low-cost camera. If there is sufficient visual data available, a machine can be trained to perform visual observations currently being done by humans. This frees humans up from routine tasks to do more abstract work and reduce or eliminate some of the human failure modes that can weaken a safety system.

The advances in machine vision algorithms have reduced the dependency on high level coding skills and extensive image labeling. These algorithms can monitor work in real time and provide active monitoring of potential hazards and work in progress. If work as performed deviates from work as planned, operators can be alerted to intervene to strengthen weakened safeguards or mitigate further drift. Long-term, this technology can be used to quantify safeguard health, improve data quality, and improve safety performance by elevating traditionally administrative safeguards into engineered safeguards. These data streams can be combined with traditional sensors and a digital twin to represent the state of a facility in real time.

This presentation will highlight the work being done by Chevron to implement this technology. Several use cases across the oil and gas value chain will be provided along with video demonstrations of the technology in use. Lessons learned from deployment and future growth of the technology will also be discussed.