Abstract
Effective risk management is critical in the gas and petroleum industry, ensuring worker safety, environmental protection, and business continuity. To improve risk management, an artificial intelligence-based system has been proposed that can identify objects and situations of risk from camera images. The system includes an alert algorithm and report generation for auditing, increasing accuracy and efficiency in oil industry operations. In addition, a no-code system has been developed, which can be customized by inserting new risk events using tools such as creating polygons and allowed and prohibited logic, and text searches to find new items in captured scenes. The system was trained by defining the objects and events to be identified, including the use of personal protective equipment, eyewash stations, handrails, fire extinguishers, prohibited areas, and vehicles traveling above the permitted speed limit. A database was created, supplemented by images from the YOLO convolutional neural network architecture, version 7, for the development of machine learning. The model was then trained, tested, and validated. The software and no-code platform were subsequently developed for the study scenario. The proposed system is scalable, versatile, and has a global territorial scope, making it suitable for various industrial monitoring situations. It can be replicated for multiple plants and customized by adding new events as required. The platform can work remotely without local servers and offers access management by registered users. The proposed system is unique, with integrated functionalities, such as vehicle speed detection, action, object, and equipment identification, and report and alert generation. The system reduces monitoring costs, is not dependent on human supervision, and can detect deviations made by employees, thus enhancing problem resolution and prevention. Finally, pre-installed cameras can be used, and the system offers customization possibilities for various industrial areas.