The Oil and Gas industry faces unique safety challenges due to its operations in harsh and remote environments. This paper presents a novel framework to address these challenges by leveraging Artificial Intelligence (AI) and Computer Vision Analytics. Our framework offers a pragmatic method for creating and implementing AI solutions that can be applied to solve similar problems and more. We have detailed the six essential stages of our framework and emphasized the challenges posed by the complexity of the Oil and Gas environment in developing and deploying these solutions.
This framework has enabled us to work with our multi-functional teams to have a structured approach and accelerate development. We were able to develop and deploy up to six use cases in a span of seven months and strengthened our ability to scale. In many of these cases we were able to build a use case from concept and increase the model’s confidence and performance significantly in just a few months. This system not only provides real-time alerts to the Health, Safety, and Environment (HSE) teams but also offers insights into historical trends of PPE usage, fostering a culture of enhanced safety through behavioral engineering. Our models are trained with industry-specific data and are integrated with existing rig infrastructure for real-time monitoring. This comprehensive approach of combining real-time safety compliance with historical trend analysis, aims to revolutionize safety practices in the Oil and Gas industry.
This paper presents the framework that was used to develop these models and highlights the challenges faced in each of these stages, the approaches taken, and shares the improvements made as a result. Although we worked on many use cases, this paper focuses on personnel and PPE detection where we have seen the highest progress in performance and adoption. With continuous testing and validation, we have seen an increase in accuracy of up to 150% from 2022 to 2023 for these use cases. Due to this framework, we were able to quickly adapt to unique challenges such as changing lighting conditions, weather conditions, cyclic setup changes due to rig-up & rig downs within these complex environments. This has led to a reduction in both false negatives and false positives. The system’s performance and scalability were validated through field tests on various rigs, further reinforcing the philosophy that the system improves as more diversified data is incorporated into the machine learning models.
The results highlight the potential of our system to greatly enhance safety practices in the Oil and Gas industry. By integrating real-time rig data with computer vision analytics, we have demonstrated the capability of AI in addressing industry-specific safety challenges. Our work contributes to the state of knowledge by showcasing the potential of advanced computer vision techniques in behavioral engineering, enhancing safety protocols, and safeguarding lives in the oil and gas industry.
